Next Article in Journal
Effect of Modified Illite on Cd Immobilization and Fertility Enhancement of Acidic Soils
Previous Article in Journal
Evolution Modes of Chili Pepper Industry Clusters under the Perspective of Social Network—An Example from Xinfu District, Xinzhou, Shanxi Province
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Sustainability
Volume 15
Issue 6
10.3390/su15064951
sustainability-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Research on a Prediction Model and Influencing Factors of Cross-Regional Price Differences of Rebar Spot Based on Long Short-Term Memory Network

by
Sen Wu
,
Shuaiqi Liu
*,
Huimin Zong
,
Yiyuan Sun
and
Wei Wang
School of Economics & Management, University of Science and Technology Beijing, No. 30 Xueyuan Road, Beijing 100083, China
*
Author to whom correspondence should be addressed.
Sustainability 2023, 15(6), 4951; https://doi.org/10.3390/su15064951
Submission received: 25 January 2023 / Revised: 15 February 2023 / Accepted: 8 March 2023 / Published: 10 March 2023
Browse Figures
Review Reports Versions Notes

Abstract

:
In this paper, taking rebar steel as an example, we study the causes and influencing factors of spot price differences in rebar steel in different regions, and put forward a prediction model of rebar steel regional price differences based on the spot price of rebar from 2013 to 2022, supply and demand, cost, macroeconomics, industrial economic indicators, and policy data. Through correlation analysis, we consider all influencing factors step by step, select indicators with high correlation to add to the model, and select the optimal combination of influencing factors by comparing the results of five groups of experiments. Using the long short-term memory network, we predict the weekly spot price differences of rebar in different regions. Based on the historical-price time series, the optimal time window setting is given as the final price difference prediction model. The experimental results show that the prediction model of rebar spot price differences can support a 72.3% effective trading rate by combining the influencing factors with the LSTM model. This study has a guiding role for spot trading and can help spot enterprises, determine arbitrage trading strategies based on the prediction results, obtain sustainable returns under low risk, and realize the maximization of cross-regional arbitrage.

1. Introduction

Promoting the cross-regional trading of steel spots is of great significance to the development of the steel industry and the improvement of steel industrial structure [1]. It can not only provide cross-regional arbitrage and risk management tools for steel production and circulation enterprises but can also provide price information for steel enterprises so they can reasonably arrange the production and help to adjust overcapacity, making steel production develop in a sustainable direction, reducing the impact of frequent price fluctuations on the smooth operation of enterprises [2].
With the rapid development of the Internet of Things, the ubiquitous connection between things and things, and things and people, can realize the intelligent perception, recognition, and management of steel production and transportation process. China Steel Net, Steel Bank, and other spot steel trading platforms have emerged, committed to creating a seamless trading platform integrating online trading, payment and settlement, warehousing and logistics, online financing, and support services [3]. Cross-regional steel spot trading is the main business of the steel spot trading platform, which undertakes consignment trading services. The platform charges commissions for part of the consignment transactions, and the charging mode is to charge service fees to the seller according to the actual sold weight, which is reflected in the bid–ask price difference. There are obvious price differences between different regions of steel, and the regional economic aggregate and the share of tradable goods are important factors that determine the regional price differences [4]. With the digital transformation of the steel industry, global steel production and transportation information can be synchronously visualized, and the prediction of price differences between regions can help companies gain huge profits.
There are three kinds of spread arbitrage, which are intertemporal arbitrage, future-spot arbitrage, and cross-regional arbitrage [5]. The so-called cross-regional arbitrage is simply a process of selling something bought in the north to the south or something sold in the east to the west to profit from the price difference of the same product in different regions [6]. The cross-regional price difference is an important index reflecting the balance of supply and demand and the macro policy of the two regions [7]. When the price determined by the supply and demand of a region becomes higher, cross-regional trade will be initiated, and transport from the region with a low price to the region with a high price will supplement the supply quantity of the region with a high price, so that the regional supply and demand will reach the equilibrium point again, and vice versa. The price difference is generally due to the different supply and demand relations between the two places [8]. A is in short supply and the price is high, while B is in oversupply and the price is low. From this point of view, to predict the price difference is to predict the market potential of the two places, so that the two places of inventory storage deployment are planned in advance. It is easier to buy at a low price, while selling depends on the downstream purchasing demand of steel. Predicting the price difference in advance allows enough of a time window to collect demand information and talk with prospective customers. Steel prices change rapidly. If the price difference is calculated based on the updated real-time quotation and then the trading transaction, it is easy to miss the arbitrage opportunity. In addition, there is usually a transit period for cross-region transactions, so it is necessary to predict the price difference in advance.
Rebar is the steel that must be used for building components above a medium size [9]. The spot market of rebar is scattered and there is information asymmetry. The transportation conditions, smelting technology, and production costs of the rebar market in different regions of China are different, so the price discovery function of the rebar futures market should reflect regional differences.
Steel price forecasting is the basis of cross-regional arbitrage, and the trend of steel prices is a time series, which reflects the changing trends of steel prices over time. In the past, there were a few pieces of research on steel price prediction. Traditional research adopted the least square method [10] and co-integration theory [11]. With the development of artificial intelligence, these methods gradually withdrew from the hot spot. The artificial intelligence method has been widely concerned. Kapl and Muller [12] proposed the traditional time series modeling method based on regression analysis but the data noise of this method affected the accuracy of the fitting. Adli [13] adopted the ARIMA model, assuming that the time series must be steady state and the error correction effect is not good. Shyu and Chang [14] used support vector regression to predict steel prices but this method is complicated and computative in price prediction, so it is difficult to popularize. The BP neural network adopted by Liu et al. [15] converges slowly in the learning process, and the network structure is difficult to determine. To overcome the above shortcomings, mixed prediction models were proposed from the perspective of algorithm improvement [16,17]; however, these models were not established in combination with the factors affecting the price change, so the results could not reflect the steel price fluctuation trend well.
As an improvement in the recurrent neural network, a long short-term memory network can well solve the time series problem and avoid long-term dependence. Many studies have verified the effectiveness and superiority of long short-term memory networks in stock price prediction based on time series data [18,19]. Similar to the stock price prediction problem, the price difference prediction of rebar spot is also a prediction problem based on time series data. The spot price difference between regions has a great relationship with historical data. The price difference of the previous period directly affects the forecast trend of the price difference of the next period.
The purpose of this study is to explore the influencing factors of steel cross-region price differences. Based on the time series of the spot price of rebar, supply and demand, cost, macro and industrial factors, policy factors, and the historical price time series, correlation analysis is carried out on the data and various factors are considered step by step. Through the correlation analysis, indicators with high correlation are selected to be added. We use the long short-term memory network to predict the spot price differences of rebar in different regions and put forward the prediction model of the spot price differences of rebar. The contributions of this paper are as follows: (1) It is the first time to select the long short-term memory network to study the prediction of steel spot price differences. (2) Based on the steel e-commerce platform as the background for the first time, we conduct data mining on the data of the steel e-commerce platform and use the data to help enterprises realize profit arbitrage. (3) Considering the regional differences in the spot price of rebar, this study puts forward an objective prediction model of the cross-regional price differences, which can assist spot enterprises to carry out cross-regional transactions to extract profits.

2. Literature Review

The existing research in this paper is mainly divided into influencing factors of price differences between regions, cross-regional arbitrage between futures and spot, and the prediction model of spot regional price differences.

2.1. Influencing Factors of Price Differences between Regions

The first is the study of the factors affecting the price differences between regions. Wu et al. [20] took rebar as an example and explored the price discovery ability of the rebar futures market based on regional differences in the spot market. It is believed that the transportation conditions, smelting technology, and production costs of the rebar market in different regions of China are different, so the sensitivity to price changes will be different. Yang et al. [21] adopted the arbitrage pricing theory to analyze the influencing factors of China’s regional carbon emission trading prices and concluded that macroeconomics, similar products, energy prices, and exchange rates would have an impact on the volatility of carbon prices.
Some of the literature does not involve the steel market but studies the price differences between regions. For example, Ke [22] believes that the price segmentation between two regions is determined by the level of economic development and the degree of trade development. Transportation cost is the main factor of the price difference between regions, and the fiscal and tax policies of local governments are the hidden factors affecting price differences. Zhang and Xie [23] proposed that factors, such as local protection and opening to the outside world, restrict the circulation of products between domestic regions and weaken the spatial transmission of prices between regions, which is an important reason for the formation of regional price differentiation. Regional output gap difference is also an important factor contributing to the price differences between regions. Some of the literature studies the influencing factors of steel market price fluctuations. For example, Li [24] pointed out the influencing factors of steel market price fluctuations include macroeconomics, steel industry reform, and trade friction. These factors are caused by the imbalance of supply and demand caused by the change in demand structure, which leads to the fluctuation in steel prices. Li et al. [25] believed that the main influencing factors of steel price fluctuations were the internal supply and demand relationship and macroeconomic fluctuation, as well as some other factors such as production cost and import and export volume. There are also some more general studies on commodity price mechanisms [26,27], most of which refer to the basic factor of supply and demand.
The above studies are the analysis of the factors affecting regional price differences. Based on the analysis of the factors affecting regional price differences of rebar, this paper combines the analysis results with the long short-term memory network to put forward the regional price difference prediction model. Compared with the above literature, this paper is a more objective and comprehensive study of the factors influencing regional price spreads and arbitrage models.

2.2. Cross-Regional Arbitrage between Futures and Spot

The price discovery mechanism has functions such as making market transactions open [28,29], enhancing cross-regional price relations [30,31], regulating market supply and demand relations [32,33], mitigating price fluctuations [34], and promoting stable and orderly development of the market [35,36]. Cross-regional arbitrage trade is the basis of price linkage between different regions [37]. Kannika and James [38] focus on the single spot market and make probabilistic forecasts for natural gas in eight spot markets. Zhou and Li [39] underscore that arbitragers play an important role in spot–futures market interaction and shock transfer, and adequate arbitrage trading during crises may help eliminate the positive basis and halt the further spread of crises. Song and Xing [40] argued that after the implementation of the continuous trading system, Shanghai copper has a better price discovery function than London copper and New York copper futures. Su et al. [41] studied the relationship between futures and the spot of crude oil with the causality analysis method. According to the research of gold futures [42], New York gold has the strongest price discovery ability, dominates the world gold price trend, and has the gold pricing power, while Shanghai gold has the weakest price discovery ability and only initially has the international gold pricing power.
It is worth noting that there are few works of literature on steel spot price discovery functions from a regional perspective. Most research on price discovery functions is based on a local or national average price of futures. Liu [43] studied the difference in price discovery function of rebar futures in different spot market regions and found that the guiding influence of the rebar spot market on futures in North and East China was significantly stronger than that in Northeast China. As for iron ore futures, it is mainly conducted from the perspective of domestic and foreign comparison. Ma [44] found the international futures price of iron ore has a significant impact on the spot price. Evren and Elif [45] found through ARCH and GARCH tests that there was a two-way guiding relationship between steel rebar futures and spot, and steel rebar futures could improve the completeness of the market. Regulators should pay attention to maintaining the connection between the raw material market and the financial market.
The above literature does not consider the cross-regional price difference and spot commodity price discovery at the same time. Based on the study of cross-regional arbitrage of spot steel, this paper proposes a price difference prediction model, which can guide enterprises to predict the steel price differences in different regions in real-time.

2.3. Prediction Model of Spot Regional Price Differences

Since spot prices are non-stationary time series in most cases, the direct use of statistical methods may lead to pseudo-regression. For the study of the relationship between non-stationary economic variables, Engle and Granger [46] proposed the co-integration theory and error correction model, which contributed a lot to the research on the short and long-term forecast of spot prices. Booth and Tse [47] found a co-integration relationship between US Treasury futures prices and Eurodollar futures prices. Elena and Eric [48] predicted short-term volatility spreads by combining volatility factors with time series. Li and Zhang [49] used the Markov transformation model to obtain the long-term co-integration relationship between Shanghai copper and London copper.
Most studies adopted a regression model to solve the problem of time series price prediction. Kapl and Adli [12,13] used the ARIMA model to predict steel price but the premise of ARIMA is that the time series must be steady state. Kim and Lim [11] used the vector error correction model and generalized autoregressive conditional heteroscedasticity method to analyze the price discovery and spillover effect of spot rebar in China. Mou et al. [50] applied a geographically weighted regression model to examine local associations between house prices and their underlying determinants, which are selected according to supply and demand in a region. Bigman et al. [51] proposed a regression of the spot price on the delivery date and the futures price at a fixed time away from the delivery date to test whether the futures price is an unbiased estimator of the spot price on the last trading day. Mehmanpazir et al. [52] used multiple logarithmic regression analyses to fit the supply and demand function and predict the price trend of Iranian steel; however, the error correction effect of the regression model is not good enough to predict the complex steel price.
With the development of artificial intelligence algorithms, scholars began to pay attention to neural networks. Alcalde et al. [53] applied an artificial neural network to predict the price of hot-rolled steel in Spain. Chou [54] developed a fuzzy time series model to predict the future trend of the trading band of the global steel price index. Ou et al. [16] proposed an extreme learning machine combined with grey correlation analysis to dynamically predict steel manufacturing costs. Liu et al. [17] adopted an adaptive neural fuzzy reasoning system to optimize the steel industry market structure from the perspective of the whole industry chain. Hsu et al. [55] proposed computational intelligence approaches based on the extended classifier system to build the inter-market arbitrage model. Haesun et al. [56] used time series analysis and modeling to study the relationship between the price of natural gas in the spot market of North America.
The prediction method of the spot price between regions has developed from the traditional cointegration statistical method to the regression model to the neural network algorithm. The LSTM algorithm used in this paper is a variant of the cyclic neural network, which is more suitable for solving the problem of steel price prediction. Different from the above literature, this paper considers the factors affecting the cross-regional price spread and the cross-regional price spread prediction. The fluctuations in steel prices are extremely complicated and essentially influenced by raw material cost, transportation cost, macroeconomics, and other factors. Therefore, it is particularly important to analyze these factors before studying the trends of steel prices. Through the correlation analysis method, we first studied the factors affecting the changes in steel prices and discussed which factors performed better in predicting steel prices. These factors were selected as input factors and the steel cross-regional price difference as output factors, and the LSTM model was established to predict the price difference, which made the prediction results more accurate and obtained a higher effective transaction rate in practice.
Through the literature research, data analysis of my steel network, and a financial research report, factors affecting the steel cross-regional price differences are obtained, as shown in Table 1.
To sum up, the existing literature has studied the influencing factors of regional price spread, cross-regional arbitrage of futures and spot, and the prediction model of spot regional price spread, while few studies have comprehensively analyzed the influencing factors of spot regional price spread and proposed a prediction model of price spread. Therefore, this paper uses correlation analysis to study the influencing factors of regional price spread, combines the influencing factors with the long short-term memory network model, and puts forward the prediction model of regional price spread of rebar to reveal the change mechanism of price spread and help enterprises realize cross-regional trade arbitrage.

3. Data Preparation

3.1. Data Selection

HRB400 rebar is one of the most common types of steel products, widely used in the civil engineering of houses, bridges, roads, and especially railways. China’s HRB400-type rebar main production areas are in North and East China. In this paper, we select HRB400 12MM rebar as the experimental object to study the spot price difference between Beijing and Shanghai rebar. Beijing and Shanghai represent North China and East China, respectively, which are important steel production and consumption centers. It is of practical significance to study the price difference between these two places. We obtained the spot price of rebar, supply and demand, cost, macroeconomic and industrial economic indicators, and policy data from April 2013 to June 2022 from a forward-looking database. The relevant indicators are shown in Table 2. With the exception of macroeconomic and industrial economic indicators, the remaining indicators all contain data from different regions.

3.2. Data Preprocessing

Some indicators in this study have data missing values. Indicators with a large proportion of missing values are removed, and the remaining ones are filled forward with data from the latest time point to fill the missing data. Most of the above missing data are concentrated on supply indicators and macroeconomic and industry indicators. Some year and month data are missing. However, nearly 70% of the supply indicators are not missing, and 76.9% of the macro and industry indicators are not missing. These indicators still could be used to study the supply factors, macroeconomics, and industry factors, and to forecast the price difference. The cross-region spread model in this study was a weekly model, so data alignment was carried out before modeling, and daily, monthly, and annual data were converted into weekly data. In addition, due to the different dimensions of each feature, the values are very different, so the maximum–minimum normalization processing is carried out on the data, and the calculation method is shown as Formula (1), where x is the sample data itself, min. is the minimum value of the sample data, and max. is the maximum value of the sample data.
z = x − min max − min
From the real data, it can be found that the spot price difference of rebar between Beijing and Shanghai fluctuates from around −400 to 400 from April 2013 to June 2022, as shown in Figure 1. The horizontal axis represents the year, and the vertical axis represents the spot price difference of rebar between Beijing and Shanghai. The volatility of the price difference between the two places is still large, which further shows the necessity of our research. In the face of severe fluctuations, forecasting the price difference can be used to better trade arbitrage and prevent risks.

4. Materials and Methods of Spot Spread Forecast

There is a price difference between regions. Buy in the regions with low prices and sell in the regions with high prices to realize arbitrage. Predicting the changes in price difference can help to know in advance when arbitrage can make profits and maximize profits. We use short-term and short-term memory networks to establish the weekly price difference model of deformed steel bars in Beijing and Shanghai. Based on the time series of rebar spot price, supply and demand, cost, macroeconomics, industry, and policy factors, the spot price difference of rebar in different regions is predicted. Based on the historical price time series, other factors are gradually considered. Through correlation analysis, the secondary index with a higher correlation is selected from the primary index that affects the cross-regional price difference, and the optimal model is selected as the final price difference prediction model.

4.1. Time Window Optimization

We use the long short-term memory network to build the weekly price difference model of rebar in Beijing and Shanghai. Based on the time series of the spot price of rebar, supply and demand, cost, macroeconomics, and industrial and policy factors, the spot price spread of rebar in different regions is predicted. Based on historical price time series, other factors are considered step by step. Through correlation analysis, the second-level index with a high correlation was selected from the first-level index affecting the cross-regional price differences, and the optimal model was selected as the final price difference prediction model.
The data is divided into training data and test data, with training data accounting for 80%. The data of the first n weeks are selected to predict the spot spread of the next week. Mean Square Error (MSE) and Mean Absolute Error (MAE) are used as evaluation indexes for the prediction results. The calculation methods are shown in Equations (2) and (3), where n is the sample quantity. The true and predicted values of a sample of y i and y i ^ , respectively.
M S E = 1 n ∑ i = 1 n ( y i − y i ^ ) 2
M A E = 1 n ∑ i = 1 n y i − y i ^
We carried out a total of five sets of experiments. The first set of experiments is only based on the historical spot price spread of rebar, and the prediction effect under different time windows is compared. The results are shown in Table 3. Combining MSE and MAE, we found that it was better to use the data of the first five weeks to predict spot spread in the sixth week. In the subsequent experiments, we set the time window as 5.

4.2. Experimental Design

After reviewing the literature, we find that the relationship between supply and demand is an important factor affecting the price difference of rebar, and the inventory level, as an intermediate index between supply and demand, is the first thing we should consider. Therefore, the second group of the experiment is based on the first group. Firstly, the inventory indexes of all places are considered, and the inventory indexes with high correlation coefficients are selected to be added to the model.
The cost of raw materials affects the supply and price of steel products. In addition, transportation costs will also have an impact on prices in the cross-regional price differences; steel and iron ore are usually shipped by sea. We also consider diesel price indicators to enrich the rating system. The third set of experiments is based on the second set of experiments by adding cost factors, including raw materials and transportation costs. Considering the balance of supply and demand, rebar is mainly used in construction materials, which is greatly affected by the demand for real estate and infrastructure. So, we put the commercial housing construction area and investment around as downstream indicators of rebar. In the fourth experiment, steel output and demand indexes are added based on the third experiment. Finally, we consider some macroeconomic indicators and industry economic indicators, which also have some impact on the spot price of rebar. In addition, some energy-saving and environmental protection policies around the region directly affect the regional steel supply and then affect the regional price differential. A fifth set of experiments incorporated these indicators. The experimental designs of the five groups are shown in Table 4.

5. Result and Discussion

The results of the five groups of experiments are shown in Table 5. In the first group of experiments, only the prediction results with a time window of five are retained. For the MAE index, to visually display the prediction effect, we carried out a restoration calculation of the normalized prediction spread, and it can be seen that the optimal MAE is about 52, which means that the absolute error of the prediction spread is about 52 yuan. The optimal weekly spread prediction model is mainly based on historical spot spread, inventory, raw materials, and transportation costs. This may be because the weekly spread model is a short-term model, sensitive to supply and demand relations and greatly affected by inventory and transportation, while macro and industrial economic factors and policies mainly affect long-term performance. We selected the model of the third group of experiments as the final weekly price difference prediction model. Figure 2 shows the spot price difference and the real price difference predicted in the test set of the third group of experiments. It can be seen that the predicted price difference can well restore the real trend and size.
All of the above indicate that the spot spread model of the Beijing and Shanghai rebar established by us is reasonable and effective, which can provide the change and trend prediction of the spread. Market traders can make decisions and participate in cross-regional spread arbitrage according to the forecast results and their own transportation and transaction costs and operating conditions. Therefore, to measure the effectiveness of trading guided by our model, this paper defines an index T i to measure effective trading and calculates the effective trading rate R. The calculation methods are shown in Equations (4) and (5), where y i and y i ^ are the real value and predicted value of spot spread, respectively, and n is the sample number of the test set, namely the total number of transactions. Simply, considering the transportation transaction cost and its own business situation, it is considered to be a valid transaction if the forecast direction is the same and the forecast price difference is not more than twice the actual price difference.
T i = 1 , y i ∗ y i ^ > 0   and   y i ^ ≤ 2 y i 0 ,   else
R = 1 n ∑ i = 1 n T i
Through calculation, the effective transaction rate of the spot spread prediction model of rebar in this paper reaches 72.3%, which indicates that the model guiding the transaction is effective.
This study points out that regional price differences are affected by historical price differences, supply and demand, cost, macroeconomics, industry, and policy. Based on this, an effective prediction model of the price difference is built by using a long short-term memory network, and the effective transaction rate reaches 72.6%. Our model forecasts the price difference of the next week based on the data of the past few weeks. Spot companies can make decisions on whether to carry out cross-regional transactions according to the predicted price difference between the two places, transportation costs, transaction costs, and time. In short, it is to compare the price differences we predict and the costs of the company’s own cross-regional transactions. If it exceeds the cost, the company can participate in cross-regional transactions and make profits.

6. Conclusions

This study is a valuable supplement to the maximization of steel spot trading profits and supports spot enterprises to reasonably choose the arbitrage area and time. Through the systematic analysis of many factors affecting the regional price difference of rebar, we obtained the key index with the greatest influence weight, which can effectively improve the accuracy of prediction. In this paper, the results of influence factor analysis are applied to the LSTM prediction model. The effective trading rate of the experimental results reveals that this method applies to the steel price prediction problem with time series characteristics, indicating that the proposed method has theoretical significance and practical application value.
Research findings: (1) Spot spread of rebar is mainly affected by inventory, raw material costs, transportation costs, and steel supply and demand, while macroeconomic and industrial economic factors and policies mainly affect long-term performance and show a poor discovery of short-term prices. (2) The optimal time window of the time series prediction model is five, that is, the price difference of the previous five weeks’ data prediction in the sixth week is set for the optimal model. (3) The spot spread prediction model proposed in this paper can promote an effective transaction rate of 72.3%, which can guide enterprises to maximize cross-regional arbitrage profits.

Author Contributions

Conceptualization, S.W. and S.L.; formal analysis, S.W., S.L., and H.Z.; methodology, S.W., S.L., H.Z., and Y.S.; validation, S.W., S.L., H.Z., Y.S., and W.W.; data curation, S.L. and H.Z.; writing—original draft preparation, S.W., S.L., H.Z., Y.S., and W.W. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Natural Science Foundation of China under Grant No. 71971025.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data in this study are obtained from the Qianyan Database at https://d.qianzhan.com/xdata/list/xCxpxwx7xY.html, accessed on 21 October 2022.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Francis, S.; Kallummal, M. India’s Comprehensive Trade Agreements: Implications for Development Trajectory. Econ. Polit. Wkly. 2013, 48, 109–122. [Google Scholar]
  2. Lin, B.; Tan, Z. Exploring arbitrage opportunities between China’s carbon markets based on statistical arbitrage pairs trading strategy. Environ. Impact Assess. Rev. 2023, 99, 107041. [Google Scholar] [CrossRef]
  3. Shen, H. Building a digital silk road? Situating the internet in China’s belt and road initiative. Int. J. Commun. 2018, 12, 19. [Google Scholar]
  4. Dreger, C.; Kosfeld, R. Do regional price levels converge? Jahrb. Nationalokon. Stat. 2010, 230, 274–286. [Google Scholar] [CrossRef]
  5. Board, J.; Sutcliffe, C. The dual listing of stock index futures: Arbitrage, Spread Arbitrage, and Currency risk. J. Futures Mark. 1996, 16, 29. [Google Scholar] [CrossRef]
  6. Olivier, M.; Albert, B. Market Power and Spatial Arbitrage between Interconnected Gas Hubs. Energy J. 2018, 39, 62. [Google Scholar]
  7. Butters, R.A.; Spulber, D.F. The extent of the market and integration through factor markets: Evidence from wholesale electricity. Econ. Inq. 2020, 58, 1076–1108. [Google Scholar] [CrossRef]
  8. Lee, C.K.M. Strategic procurement from forward contract and spot market. Ind. Manag. Data Syst. 2014, 114, 778–796. [Google Scholar] [CrossRef]
  9. Kim, S.; Yun, D.; Park, W.; Jang, Y.I. Bond strength prediction for deformed steel rebar embedded in recycled coarse aggregate concrete. Mater. Des. 2015, 83, 257–269. [Google Scholar] [CrossRef]
  10. Li, M.; Sun, H.; Agyeman, F.O. Analysis of Potential Factors Influencing China’s Regional Sustainable Economic Growth. Appl. Sci. 2021, 11, 10832. [Google Scholar] [CrossRef]
  11. Kim, K.; Lim, S. Price discovery and volatility spillover in spot and futures markets: Evidences from steel-related commodities in China. Appl. Econ. Lett. 2019, 26, 351–357. [Google Scholar] [CrossRef]
  12. Kapl, M.; Muller, W. Prediction of steel prices: A comparison between a conventional regression model and MSSA. Statist. Interface 2010, 3, 369–375. [Google Scholar] [CrossRef] [Green Version]
  13. Adli, K. Forecasting steel prices using ARIMAX model: A case study of Turkey. Int. J. Bus. Manag. Technol. 2020, 12, 70–87. [Google Scholar]
  14. Shyu, Y.; Chang, C. A hybrid model of MEMD and PSO-LSSVR for steel price forecasting. Int. J. Eng. Manag. Res. 2022, 12, 30–40. [Google Scholar] [CrossRef]
  15. Liu, Z.; Wang, Y.; Zhu, S.; Zhang, B.; Wei, L. Steel prices index prediction in China based on BP neural network. In Proceedings of 4th International Conference on Logistics, Informatics and Service Science, LISS 2014, Berkeley, CA, USA, 23 July 2014; Springer: Berlin/Heidelberg, Germany, 2015; pp. 603–608. [Google Scholar]
  16. Ou, T.; Cheng, C.; Chen, P.; Perng, C. Dynamic cost forecasting model based on extreme learning machine-A case study in steel plant. Comput. Ind. Eng. 2016, 101, 544–553. [Google Scholar] [CrossRef]
  17. Liu, Y.; Li, H.; Guan, J.; Liu, X.; Guan, Q.; Sun, Q. Influence of different factors on prices of upstream, middle and downstream products in China’s whole steel industry chain: Based on Adaptive Neural Fuzzy Inference System. Resour. Policy 2019, 60, 134–142. [Google Scholar] [CrossRef]
  18. Xu, Y.; Wang, Z.; Wu, Z. A CNN-BiLSTM-based Multi-feature Integration Model for Stock Trend Prediction. Data Anal. Knowl. Discov. 2021, 5, 126–137. [Google Scholar]
  19. Qi, T.; Jiang, H. Exploring Stock Price Trend Using Seq2Seq Based Automatic Text Summarization and Sentiment Mini. Manag. Rev. 2021, 33, 257–269. [Google Scholar]
  20. Wu, G.H.; Jiang, X.L.; Deng, S.Y. A Characteristic Analysis and Countermeasure Study of the Hedging of Listed Companies in China Stock Markets. J. Asian Finan. Econ. Bus 2021, 8, 147–158. [Google Scholar]
  21. Yang, X.; Zhang, C.; Yang, Y.; Wu, Y.; Yun, P.; Ali, W. China’s Carbon Pricing Based on Heterogeneous Tail Distribution. Sustainability 2020, 12, 2754. [Google Scholar] [CrossRef] [Green Version]
  22. Ke, S. Domestic market integration and regional economic growth—China’s recent experience from 1995–2011. World Dev. 2015, 66, 588–597. [Google Scholar] [CrossRef]
  23. Zhang, M.; Xie, J. Spatial Correlation of Regional Prices and Influencing Factors of Transmission Differences in China: An Empirical Study Based on Dynamic Spatial Panel Model. J. Financ. Econ. 2012, 38, 12. [Google Scholar]
  24. Li, J. Analysis of China’s Economic Downturn: From the Perspective of Macro-control. In Proceedings of the 7th International Conference on Financial Innovation and Economic Development (ICFIED 2022), Harbin, China, 21–23 January 2022; Atlantis Press: Paris, France, 2022; pp. 988–994. [Google Scholar]
  25. Li, H.; Wang, B.; Ren, E.; Wu, C. Empirical analysis of the influencing factors on iron ore prices. In Proceedings of the 2nd International Conference on Artificial Intelligence, Management Science and Electronic Commerce (AIMSEC), Zhengzhou, China, 8–10 August 2011; pp. 3004–3008. [Google Scholar]
  26. Wang, H.; Sheng, H.; Zhang, H. Influence factors of international gold futures price volatility. Trans. Nonferr. Metals Soc. China 2019, 29, 2447–2454. [Google Scholar] [CrossRef]
  27. Sverdrup, H.U.; Ragnarsdottir, K.V. A system dynamics model for platinum group metal supply, market price, depletion of extractable amounts, ore grade, recycling and stocks-in-use. Resour. Conserv. Recycl. 2016, 114, 130–152. [Google Scholar] [CrossRef]
  28. Zeng, Y.; Dong, C.; Mikael, H.; Sun, J.; Shi, D. Can the Shanghai LNG Price Index indicate Chinese market? An econometric investigation using price discovery theory. Front. Energy 2020, 14, 726–739. [Google Scholar] [CrossRef]
  29. Cao, C.; Ghysels, E.; Hatheway, F. Price discovery without trading: Evidence from the Nasdaq preopening. J. Financ. 2000, 55, 1339–1365. [Google Scholar] [CrossRef]
  30. Cudjoe, G.; Breisinger, C.; Diao, X. Local impacts of a global crisis: Food price transmission, consumer welfare and poverty in Ghana. Food Policy 2010, 35, 294–302. [Google Scholar] [CrossRef]
  31. Popatl, M.; Griffith, G.; Mounter, S.; Cacho, O. Infrastructure investments, regional trade agreements and agricultural market integration in Mozambique. Food Security 2021, 14, 9–22. [Google Scholar] [CrossRef]
  32. Figueiredo, N.C.; da Silva, P.P.; Bunn, D. Weather and market specificities in the regional transmission of renewable energy price effects. Energy 2016, 114, 188–200. [Google Scholar] [CrossRef] [Green Version]
  33. Shi, Y.; Wu, F.; Chu, L.K.; Sculli, D.; Xu, Y.H. A portfolio approach to managing procurement risk using multi-stage stochastic programming. J. Operat. Res. Soc. 2011, 62, 1958–1970. [Google Scholar] [CrossRef]
  34. Willems, B.; Kupper, G. Arbitrage in Energy Markets: Price Discrimination under Congestion. Energy J. 2010, 31, 41–66. [Google Scholar] [CrossRef] [Green Version]
  35. Rui, X.; Feng, L.; Feng, J. A gas-on-gas competition trading mechanism based on cooperative game models in China’s gas market. Energy Rep. 2020, 6, 365–377. [Google Scholar] [CrossRef]
  36. Kumar, S. Price discovery and market efficiency: Evidence from agricultural commodities futures markets. S. Asian J. Manag. 2004, 11, 32. [Google Scholar]
  37. Jiang, M.; An, H.; Jia, X.; Sun, X. The influence of global benchmark oil prices on the regional oil spot market in multi-period evolution. Energy 2016, 118, 742–752. [Google Scholar] [CrossRef]
  38. Kannika, D.; James, W. Prequential forecasting in the presence of structure breaks in natural gas spot markets. Empir. Econ. 2019, 59, 2363–2384. [Google Scholar]
  39. Zhou, X.; Li, M. Shock Transfer in Futures and Spot Markets: An Agent-Based Simulation Modelling Method. Discr. Dyn. Nat. Soc. 2021, 2021, 7386169. [Google Scholar] [CrossRef]
  40. Hua, R.; Lu, B.; Chen, B. Price discovery process in the copper markets: Is Shanghai futures market relevant? Rev. Futur. Mark. 2010, 18, 299–312. [Google Scholar]
  41. Su, X.; Zhu, H.; Yang, X. Heterogeneous Causal Relationships between Spot and Futures Oil Prices: Evidence from Quantile Causality Analysis. Sustainability 2019, 11, 1359. [Google Scholar] [CrossRef] [Green Version]
  42. Xu, C.; Zhang, D. Market openness and market quality in gold markets. J. Futur. Mark. 2019, 39, 384–401. [Google Scholar] [CrossRef]
  43. Liu, G. Technical trading behaviour: Evidence from Chinese rebar futures market. Comput. Econ. 2019, 54, 669–704. [Google Scholar] [CrossRef] [Green Version]
  44. Ma, Y. Iron ore spot price volatility and change in forward pricing mechanism. Resour. Policy 2013, 38, 621–627. [Google Scholar] [CrossRef] [Green Version]
  45. Evren, A.; Elif, M. Chinese steel market in the post-futures period. Resour. Policy 2014, 42, 10–17. [Google Scholar]
  46. Engle, R.F.; Granger, C.W.J. Co-Integration and Error Correction: Representation, Estimation, and Testing. Econom. J. Economet. 1987, 55, 251–276. [Google Scholar] [CrossRef]
  47. Booth, G.G.; Tse, Y. Long Memory in Interest Rate Futures Markets: A Fractional Cointegration Analysis. J. Futur. Mark. 1995, 15, 573–584. [Google Scholar] [CrossRef]
  48. Andreou, E.; Ghysels, E. Predicting the VIX and the volatility risk premium: The role of short-run funding spreads Volatility Factors. J. Economet. 2020, 220, 366–398. [Google Scholar] [CrossRef]
  49. Li, X.D.; Zhang, B. Price discovery for copper futures in informationally linked markets. Appl. Econ. Lett. 2009, 16, 1555–1558. [Google Scholar] [CrossRef]
  50. Mou, Y.; He, Q.; Zhou, B. Detecting the spatially non-stationary relationships between housing price and its determinants in China: Guide for housing market sustainability. Sustainability 2017, 9, 1826. [Google Scholar] [CrossRef] [Green Version]
  51. Bigman, D.; Goldfarb, D.; Schechtman, E. Futures Market Efficiency and the Time Content of the Information Sets. J. Futur. Mark. 1983, 3, 321–334. [Google Scholar] [CrossRef]
  52. Mehmanpazir, F.; Khalili-Damghani, K.; Hafezalkotob, A. Modeling steel supply and demand functions using logarithmic multiple regression analysis (case study: Steel industry in Iran). Resour. Policy 2019, 63, 101409. [Google Scholar] [CrossRef]
  53. Alcalde, R.; Urda, D.; Armino, C.; García, S.; Manzanedo, M.; Herrero, Á. Non-linear Neural Models to Predict HRC Steel Price in Spain. In Proceedings of the 17th International Conference on Soft Computing Models in Industrial and Environmental Applications (SOCO 2022), Salamanca, Spain, 5–7 September 2022; Springer Nature: Cham, Switzerland, 2022; pp. 186–194. [Google Scholar]
  54. Chou, M. An application of fuzzy time series: A long range forecasting method in the global steel price index forecast. Rev. Econ. Financ. 2013, 3, 90–98. [Google Scholar]
  55. Hsu, Y.-C.; Chen, A.-P.; Chang, J.-H. An inter-market arbitrage trading system based on extended classifier systems. Expert Syst. Appl. 2011, 38, 3784–3792. [Google Scholar] [CrossRef]
  56. Park, H.; James, W.; Bessler, D.A. Price interactions and discovery among natural gas spot markets in North America. Energy Policy 2007, 36, 290–302. [Google Scholar] [CrossRef]
Sustainability 15 04951 g001 550
Figure 1. The spot price difference of rebar in Beijing and Shanghai.
Figure 1. The spot price difference of rebar in Beijing and Shanghai.
Sustainability 15 04951 g001
Sustainability 15 04951 g002 550
Figure 2. Forecast results of spot spread.
Figure 2. Forecast results of spot spread.
Sustainability 15 04951 g002
Table
Table 1. Summary of influencing factors of steel cross-region price differences.
Table 1. Summary of influencing factors of steel cross-region price differences.
Primary FactorsSecondary Factors
MacroscopicalMacroeconomic factors (liquidity indicators, macroeconomic indicators, inflation indicators, exchange rates)
SupplyInventory level, suspension, and resumption of steel production, inflow of foreign resources, outflow of own resources
PolicyAir pollution prevention and control regulations, environmental supervision, energy control, financial subsidies, production restriction policies
DemandReal-estate projects and infrastructure projects (investment), terminal industrial clusters (sales), market trading sentiment
CostFreight, raw material price, energy consumption factors
TechnologyProduction and smelting technology
EconomyEconomic growth rate, the return of funds from steel mills
ClimateClimate impacts the use and storage costs; climate affects construction and therefore demand
MarketThe number of steel traders
Table
Table 2. Description of experimental data indexes.
Table 2. Description of experimental data indexes.
CategoryIndexRemarks
Supply and demand RelationshipRebar inventorySupply, weekly data
Steel outputSupply, monthly data
Construction area of commercial housingDemand, monthly data
New commercial housing construction areaDemand, monthly data
Investment in commercial housing developmentDemand, monthly data
CostDomestic coastal freight of iron oreFreight, daily data
Wholesale diesel priceFreight, daily data
Steelmaking pig iron priceRaw materials, daily data
Secondary metallurgical coke priceRaw materials, daily data
Scrap priceRaw materials, daily data
Macroeconomics and industryChina Commodity Price Index—SteelWeekly data
Ferrous metal smelting and rolling industry: Cost of products sold/net accounts receivable/number of business unitsMonthly data
Consumer price indexMonthly data
Macroeconomic climate indexMonthly data
PolicyGovernment expenditure—energy conservation and environmental protectionAnnual data
Spot priceRebar spot priceDaily data
Table
Table 3. The forecast results only consider spot prices under different time windows.
Table 3. The forecast results only consider spot prices under different time windows.
Time Window12345
MSE0.6060.5590.5550.5460.543
MAE0.0630.0590.0590.0600.060
Table
Table 4. Index selection of the five groups of experiments.
Table 4. Index selection of the five groups of experiments.
ExperimentIndex
Experiment 1Cross-regional price differences
Experiment 2Cross-regional price differences, Inventory
Experiment 3Cross-regional price differences, Inventory, Transportation costs
Experiment 4Cross-regional price differences, Inventory, Transportation costs, Steel production and demand
Experiment 5Cross-regional price differences, Inventory, Transportation costs, Steel production and demand, Macroeconomic indicators, Industry index
Table
Table 5. The prediction model of weekly price spread considering different indices.
Table 5. The prediction model of weekly price spread considering different indices.
Experiment12345
MSE0.5430.5230.5190.5250.647
Antinormalized MAE54.60753.41952.42852.22057.928
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Wu, S.; Liu, S.; Zong, H.; Sun, Y.; Wang, W. Research on a Prediction Model and Influencing Factors of Cross-Regional Price Differences of Rebar Spot Based on Long Short-Term Memory Network. Sustainability 2023, 15, 4951. https://doi.org/10.3390/su15064951

AMA Style

Wu S, Liu S, Zong H, Sun Y, Wang W. Research on a Prediction Model and Influencing Factors of Cross-Regional Price Differences of Rebar Spot Based on Long Short-Term Memory Network. Sustainability. 2023; 15(6):4951. https://doi.org/10.3390/su15064951

Chicago/Turabian Style

Wu, Sen, Shuaiqi Liu, Huimin Zong, Yiyuan Sun, and Wei Wang. 2023. "Research on a Prediction Model and Influencing Factors of Cross-Regional Price Differences of Rebar Spot Based on Long Short-Term Memory Network" Sustainability 15, no. 6: 4951. https://doi.org/10.3390/su15064951

APA Style

Wu, S., Liu, S., Zong, H., Sun, Y., & Wang, W. (2023). Research on a Prediction Model and Influencing Factors of Cross-Regional Price Differences of Rebar Spot Based on Long Short-Term Memory Network. Sustainability, 15(6), 4951. https://doi.org/10.3390/su15064951

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop