Prediction and Finding Spam in Social Network Using Convolutional Neural Network Over Ridge Linear Classification to Improve Precision Rate

Prediction and Finding Spam in Social Network Using Convolutional Neural Network Over Ridge Linear Classification to Improve Precision Rate

연구 분야: Artificial Intelligence

학회: 2024 Ninth International Conference on Science Technology Engineering and Mathematics (ICONSTEM)

The suggested research would build a system for detecting data intrusions in order to stop data theft by contrasting the general performance of two distinct machine learning (ML) approaches. Neural networks, specifically the Novel Convolutional Neural Network and ridge linear regression classifiers, are selected for this study. Supplies and Procedures: This procedure includes selecting the gathering of data, training it, and testing it using suggested classifiers. The results of two classifiers are divided into two groups for SPSS analysis, with ten iteration values under various functional operations in each group; finally it counts 20 iterations, Gpower at 80% (g power parameters: =0.05, power =0.80). The study's sample size was 1672 and sourced from the kaggle website. The experimental analysis is carried out using python compiler and the accuracy gain of both classifiers are compared. The selected Novel Convolutional Neural Network classifier improved security to avoid data theft identification with the accuracy of 91.23% and ridge regression classifier gained accuracy of 90.11%. The study has the no significance value of \mathrm{p}=0.517 (\mathrm{p} > 0.05) . Conclusion: The accuracy score of the Ridge Linear Classification is 90.11%, whereas the accuracy score of the Convolutional Neural Network is 91.23%. The analysis demonstrates that Ridge Linear Classification (90.11%) is outperformed by Convolutional Neural Network (91.23%).