INFLUENCEAI: A DEEP LEARNING FRAMEWORK FOR REALTIME SOCIAL MEDIA INFLUENCER SCORING

Abstract

The AI-based Social Network Influencer Analyzer used to evaluate influencer impact through engagement and authenticity metrics, addressing inefficiencies in traditional influencer analysis. Existing manual methods for influencer analysis are time-consuming, requiring analysts to manually collect and evaluate metrics like followers, likes, and engagement rates, often leading to inconsistent results and scalability issues due to the subjective nature of human judgment. The proposed method introduces an automated system built with Python and Tkinter, featuring a dual-role GUI for admins and users; admins can upload datasets, preprocess them using MinMaxScaler, and train models (SVR, KNN, and CNN 1D Regressor) to predict Influence_Score, while users can generate predictions on new data and compare model performances via an HTML table. The CNN 1D Regressor, with its architecture of four convolutional layers (128, 64, 64, 32 filters), max pooling, and dense layers (32, 16, 1 neurons), effectively captures complex patterns in the data, providing a scalable and accurate solution for identifying impactful influencers in social media marketing. The system processes a dataset with 1000 records and 8 features, achieving high predictive accuracy with the CNN 1D Regressor model (R²: 0.9957, MAE: 0.1330, RMSE: 0.1965), outperforming SVR (R²: 0.9830, MAE: 0.3219, RMSE: 0.3886) and KNN Regressor (R²: 0.9751, MAE: 0.3776, RMSE: 0.4712)