Hybrid Cascade Forward Neural Network with Elman Neural Network for Disease Prediction

Software Defect Prediction through Neural Network and Feature Selections

Software failure such as software defect causes billion of dollar loss every year. Software failure also affects billion of people worldwide. Inadequate software testing can cause software failure. To predict the software defect, this study proposed a model consisting of feature selection and classifications. The correlation base method was used for feature selection, and radial base function neural network (RBF) was used for classification. Also, for testing the proposed system, fourteen NASA data sets were used including CM1, JM1, KC1, KC2, KC3, KC4, MC1, MC2, MW1, PC1, PC2, PC3, PC4, and PC5. The data set was divided using the well-known K-cross-validation methods which were performed to divide the data set for training and testing the RBF. The RBF were trained and tested before and after feature selections. Precision, recall, F-measure, and accuracy are four methods used to evaluate the performance of the proposed methods. The precision obtained for the fourteen data sets was CM1, 94.01%; JM1, 85.18%; KC1, 83.24%; KC2, 81.27%; KC3, 79.30%; KC4, 85.29%; MC1, 99.89%; MC2, 73.27%; MW1, 90.90%; PC1, 98.79%; PC2, 100%; PC3, 95.67%; PC4, 95.12%; and PC5, 80.89%. Recall was as follows: CM1, 95.78%; JM1, 87.89%; KC1, 86.24%; KC2, 83.82%; KC3, 82.10%; KC4, 86.28%; MC1, 100%; MC2, 76.67%; MW1, 92.09%; PC1, 99.98%; PC2, 100%; PC3, 96.23%; PC4, 95.17%; and PC5, 81.80%. F-measure was as follows: CM1, 0.95; JM1, 0.87; KC1, 0.83; KC2, 0.82; KC3, 0.85; KC4, 0.86; MC1, 0.99; MC2, 0.76; MW1, 0.95; PC1, 0.99; PC2, 0.99; PC3, 0.97; PC4, 0.95; and PC5, 0.80. The accuracy obtained was as follows: CM1, 93.99%; JM1, 84.87%; KC1, 83.25%; KC2, 79.11%; KC3, 78.25%; KC4, 83.18%; MC1, 99.01%; MC2, 70.18%; MW1, 88.90%; PC1, 98.99%; PC2, 99.80%; PC3, 94.11%; PC4, 94.4%; and PC5, 79.02%. The proposed method results were compared with the result obtained from different methods. The proposed model obtained better results than other methods for data set CM1, KC4, MC1, PC1, PC2, PC3, PC4, and PC5.

Evaluation of a cascade artificial neural network for modeling and optimization of process parameters in co-composting of cattle manure and municipal solid waste

The present study was carried out to improve, test, and validate the Cascade Forward Neural Network (CFNN) for co-composting of municipal solid waste (MSW) and cattle manure (CM). Composting was performed in vessel pilot-scale reactors with different CM rates for 105 days. The CFNN used 5 input variables containing CM and MSW mixture combinations, and 1 output for each of the compost quality parameters. The CFNN results were compared with Response Surface Methodology (RSM) and Feed Forward Neural Network (FFNN) results. Multi-objective optimization process using Genetic Algorithm (GA), the total desirability, which has a much better value than the RSM, was obtained as 0.4455 and the CM ratio and processing time were determined as approximately 23.39% and 104.86 days, respectively. It is concluded that CFNN is a unique modeling tool, exhibiting superior modeling and prediction performance in MSW and compost modeling for CM.

Predictive study of tuberculosis incidence by time series method and Elman neural network in Kashgar, China

OBJECTIVES

Kashgar, located in Xinjiang, China has a high incidence of tuberculosis (TB) making prevention and control extremely difficult. In addition, there have been very few prediction studies on TB incidence here. We; therefore, considered it a high priority to do prediction analysis of TB incidence in Kashgar, and so provide a scientific reference for eventual prevention and control.

DESIGN

Time series study.

SETTING KASHGAR, CHINA

Kashgar, China.

METHODS

We used a single Box-Jenkins method and a Box-Jenkins and Elman neural network (ElmanNN) hybrid method to do prediction analysis of TB incidence in Kashgar. Root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) were used to measure the prediction accuracy.

RESULTS

After careful analysis, the single autoregression (AR) (1, 2, 8) model and the AR (1, 2, 8)-ElmanNN (AR-Elman) hybrid model were established, and the optimal neurons value of the AR-Elman hybrid model is 6. In the fitting dataset, the RMSE, MAE and MAPE were 6.15, 4.33 and 0.2858, respectively, for the AR (1, 2, 8) model, and 3.78, 3.38 and 0.1837, respectively, for the AR-Elman hybrid model. In the forecasting dataset, the RMSE, MAE and MAPE were 10.88, 8.75 and 0.2029, respectively, for the AR (1, 2, 8) model, and 8.86, 7.29 and 0.2006, respectively, for the AR-Elman hybrid model.

CONCLUSIONS

Both the single AR (1, 2, 8) model and the AR-Elman model could be used to predict the TB incidence in Kashgar, but the modelling and validation scale-dependent measures (RMSE, MAE and MAPE) in the AR (1, 2, 8) model were inferior to those in the AR-Elman hybrid model, which indicated that the AR-Elman hybrid model was better than the AR (1, 2, 8) model. The Box-Jenkins and ElmanNN hybrid method therefore can be highlighted in predicting the temporal trends of TB incidence in Kashgar, which may act as the potential for far-reaching implications for prevention and control of TB.

A comparative study of two methods to predict the incidence of hepatitis B in Guangxi, China

In recent years, the incidence of hepatitis B (HB) in Guangxi is higher than that of the national level; it has been increasing, so it is urgent to do a good predictive research of HB incidence, which can help analyze the early warning of hepatitis B in Guangxi, China. In the study, the feasibility of predicting HB incidence in Guangxi by autoregressive integrated moving average (ARIMA) model method and Elman neural network (ElmanNN) method was discussed respectively, and the prediction accuracy of the two models was compared. Finally, we established the ARIMA (0, 1, 1) model and ElmanNN with 8 neurons. Both ARIMA (0, 1, 1) model and ElmanNN model had good performance, and their prediction accuracy were high. The fitting and prediction root-mean-square error (RMSE) and mean absolute error (MAE) of ElmanNN were smaller than those of ARIMA (0, 1, 1) model, which indicated that ElmanNN was superior to ARIMA (0, 1, 1) model in predicting the incidence of hepatitis B in Guangxi. Based on the ElmanNN, the HB incidence from September 2019 to December 2020 in Guangxi was predicted, the predicted results showed that the incidence of HB in 2020 was slightly higher than that in 2019 and the change trend was similar to that in 2019, for 2021 and beyond, the ElmanNN model could be used to continue the predictive analysis.

Prediction of intrapartum fetal hypoxia considering feature selection algorithms and machine learning models

Introduction

Cardiotocography (CTG) consists of two biophysical signals that are fetal heart rate (FHR) and uterine contraction (UC). In this research area, the computerized systems are usually utilized to provide more objective and repeatable results.

Materials and Methods

Feature selection algorithms are of great importance regarding the computerized systems to not only reduce the dimension of feature set but also to reveal the most relevant features without losing too much information. In this paper, three filters and two wrappers feature selection methods and machine learning models, which are artificial neural network (ANN), k-nearest neighbor (kNN), decision tree (DT), and support vector machine (SVM), are evaluated on a high dimensional feature set obtained from an open-access CTU-UHB intrapartum CTG database. The signals are divided into two classes as normal and hypoxic considering umbilical artery pH value (pH < 7.20) measured after delivery. A comprehensive diagnostic feature set forming the features obtained from morphological, linear, nonlinear, time-frequency and image-based time-frequency domains is generated first. Then, combinations of the feature selection algorithms and machine learning models are evaluated to achieve the most effective features as well as high classification performance.

Results

The experimental results show that it is possible to achieve better classification performance using lower dimensional feature set that comprises of more related features, instead of the high-dimensional feature set. The most informative feature subset was generated by considering the frequency of selection of the features by feature selection algorithms. As a result, the most efficient results were produced by selected only 12 relevant features instead of a full feature set consisting of 30 diagnostic indices and SVM model. Sensitivity and specificity were achieved as 77.40% and 93.86%, respectively.

Conclusion

Consequently, the evaluation of multiple feature selection algorithms resulted in achieving the best results.