A robust hybrid methodology between applied linear regression model (alrm) and multilayer perceptron (mlp)

Abstract

Background: The goal of this study is to illustrate an optimum variable selection method using established Multiple Linear Regression (MLR) models and to validate the variable using Multilayer Perceptron Neural Network (MLP) models. Initially, all selected variables will be passed through the bootstrap methodology, and they were screened for significant relationships.

Objective: The goal of this work is to analyze and construct a model for the factor linked with total crime cases by combining an Applied Linear Regression Model (ALRM) and a Multilayer Perceptron (MLP).

Material and Methods: Around 200 data was simulated to build the methodology. Advanced computational statistical modeling methodologies were used to evaluate data descriptions of several variables in this retrospective study, including the total victim, gender, age, marital status, social class, adult in the household, children in household, burglary’s victim, sexual’s victim, victim’s report, and household location. The case study was developed and implemented using the R-Studio program and syntax.

Results: The statistical method demonstrated that regression modeling surpasses R-squared and mean square error test in most situations. Researchers observed that when data is divided into two datasets for training and testing, the hybrid model approach performs significantly better at predicting the experiment’s outcome. When it came time to determine variable validity, the well-established bootstrap-integrated MLR approach was applied. Ten characteristics are taken into consideration in this case: Gender (: -0.4369700; p< 0.25), age (: -0.0086757; p< 0.25), marital status (: 0.2646097; p< 0.25), social class ( : 0.0602540; p< 0.25), adult in household (: -0.0211293; p> 0.25), children in household (: -0.0025346; p> 0.25), burglary’s victim (: 1.3473593; p< 0.25), sexual’s victim (: 1.0382444; p< 0.25), victim’s report (: -0.3176104; p< 0.25), and location of household (: -0.1355046; p< 0.25).There is a 0.07745823 MSE for the linear model in this scenario.

Conclusion: The neural network’s Predicted Mean Square Error (PMSE) was used to assess MLP’s performance (MSE-forecasts the Network). PMSE is used to determine how far our projections are from the actual data, and the lowest MSE from the MLP indicates the best achievement. The R syntax for MLR and MLP is also included in this research article.As a result, the study’s conclusion establishes the superiority of the hybrid model technique.

Bangladesh Journal of Medical Science Vol. 22 No. 01 January’23 Page : 38-46