A Novel Hybrid Model for Short-Term Traffic Flow Prediction Based on Extreme Learning Machine and Improved Kernel Density Estimation

Round 1

Reviewer 1 Report

 Thank's for your work. You could provide your collected datas to communutiy as well, but this is not necessary

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper proposed a model for improving the accuracy of short-term traffic flow prediction based on ELM, AKDE, and CKDE. ELM was employed to handle non-linear prediction, AKDE was adopted to estimate the bandwidth of CKDE, and CKDE was applied to handle the residuals obtained by ELM. The performance of the proposed model was evaluated on two datasets, the results indicated that the proposed model obtained better performance than other concerned models.

There are some problems need to be solved, as shown below.

(1) Expressions should be improved.

Some sentences contain grammatical mistakes or are not complete sentences, such as, “Divide ...” and “Establish ...” in section 2.3, the title “Improved percentages...” of Table 3, and so on.

(2) Formulas should be checked again.

For example, in formula (9)，X=[x₁, ... ,x_N]R^(n-d)xd，the detailed matrix of X seems wrong.

(3) Experiments setting should be provided.

We cannot find the parameters of ARIMA in this paper. In fact, different parameters may have great impacts on prediction results.

(4) About the evaluation metrics.

We suggest that the metrics should be named more professionally. What's the difference between MRPE and MAPE? What's the difference between RMSRE and RMSPE?

(5) About Baselines.

In conventional models, only ARIMA is taken for comparison, more baselines should be added.

(6) About the advantages of the proposed model.

As we know, there are different deep learning (DL) models for traffic flow prediction, what are the advantages of your proposed model?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The title should not have unknown or many abbreviations. So, please elaborate ELM-AKDE-CKDE.

Fig. 2: What is the meaning of the dashed blocks? What are the inputs to the right-side dashed block?

What is the dimension of the input vector? What does it correspond?

Fig. 7: What is the unit of 'Time Series'?

The proposed method should be compared with state-of-the-art methods with references.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

In Figure 3, the detailed address of the intersection is needed.

Reviewer 3 Report

Thank you for addressing the comments.