An Innovation Success Prediction Model of Android Application Using Logistic Regression Over MLC in Combination with PCA

Abstract

The goal of this work is to assess the correctness and exactness of LR and Maximum Likelihood Classification (MLC) Classification algorithms in predicting the success of Android applications. A framework for predicting the success rate of Android applications that compares Logistic Regression and Maximum Likelihood classifiers has been proposed and developed. The sample size was determined using G powers to be 10 in each category. Sample size was calculated using clinical analysis, with alpha and beta numbers of 0.05 and 0.5, 95% assurance, and 80% well before power. The following results are obtained by running algorithms for various iterations. The Logistic Regression classifier predicts the success rate of an Android application with an accuracy of 80.3%, while the Maximum Likelihood classifier predicts it with 95.1%. The significance level is 0.001 $(\mathbf{p}\mathbf{0.005})$. As a result, Maximum Likelihood Classification outperforms LR classifiers. In terms of precision and accuracy, the results show that the Maximum Likelihood classification (MLC) outperforms the Logistic Regression.