*3.1. Measurements*

Initially, the PART rule classifier was tested on the dataset to measure the classification accuracy with the seed of random numbers selected for XVal. The percentage was 1, the confidence factor was 0.25%, the minimum number of objects was 2, and the number of folds was set to 3. After loop tests, the average accuracy of the final result was 99.28%. Secondly, the same measurement was tested on the Decision table rule classifier. The final result with an average accuracy of 98.22% was obtained in 0.77 s. The subset value was 99.60%, and the average error was 0.03%. By employing the rule classification (PART and Decision table), good predictive rules were obtained for the patient's care. The outcomes in the initial phase were the most appropriate with a mean accuracy of 98.75%; the error rate remained at 0.02%.

The results obtained for the classification accuracy are presented in Table 1 along with the attribute details and the clustering instances for the classification. It is comprised of three sections. The first section discusses the details of the properties used for the Weka platform for assessment, with 281 patients describing their age limits by classification type and improving the evaluation of positive

and negative tested weights. Additionally, it provides accurate information and average classification accuracy for PART and Decision table rule classifiers, including kappa statistics, mean error, true positive rate, false positive rate, accuracy, recall rate, F-measure, Matthew's correlation curve (MCC)., Receiver operating characteristics (ROC), Precision recall curve (PRC) area ratios, and the time it takes for a prediction analysis [39,40].


• From age >40 and ≤60 = 144 patients From>60and≤8076patients

=

 age •Fromage >80 =1patient

•


1 NID = not insulin dependent; GTD = gestational diabetes patients; IND = insulin dependent; MCC = Matthew's correlation curve; ROC = Receiver operating characteristics; PRC = Precision recall curve; N = number of patients; ≥greater than; ≤less than; % = percentage value; T\_N = tested negative; T\_P = tested positive; Values = two clusters 0 and 1; N´ = total number of classified patients.

The details of the cluster instance, as shown in Figure 4, was tested and classified as positive/negative. Out of 281 instances, 138 (49.11%) were classified as the 0 cluster instance, among them 47 (16.72%) were tested as negative, and 91 (32.38%) were tested as positive. One hundred and forty-three (50.88%) were classified as a cluster 1 instance from which 40 (14.23%) were tested as negative and 103 (36.65%) were tested as positive. In the final assessment, 51% were classified as positive and 49% instances as negative. The values of these classifications were used as input to the regression prediction phase.

**Figure 4.** Evaluation of the Kmean clusters tested as positive and negative.

#### *3.2. Rule Forecast Assessment*

The predictive analysis represents the assessment for decision-making by determining the ratio of patient characteristics. The forecast analysis obtained in the study is graphically displayed in Figure 5a–g, and the 23 rules achieved through the rule classification measurements are described in Table 2 in terms of the patients' initial screening stage of healthcare.

**Figure 5.** *Cont.*

(**c**) 

(**d**) **Figure 5.** *Cont.*

(**f**) 

(**g**) 

**Figure 5. (a–g)** The regression prediction assessment of the seven main features used for the analysis of clinical significance.


**Table 2.** Twenty-three if-then rules achieved from the classification analysis.


**Table 2.** *Cont.*

The prediction assessment by logistic regression used in this study for clinical significance was analyzed by the confidence interval of 0.95%. The patient features used were age, blood glucose, body mass index, physical exercise, family history of diabetes, family cardiovascular history, and work stress by the M5 method in regression. The results of the forecast prediction for diabetes mellitus patients on the age feature show that patients up to 51 years could have a high death risk if the ratio of other features include a glucose level of 120.45 mmol/L, BMI ≥ 23, physical exercise between 0.5 to 0.6, family diabetes history of 0.6, cardiovascular stroke history of 0.61, and a work-stress ratio count of 1.08.
