Machine Learning-Based Prediction of User Activity on Instagram: Identifying Active and Inactive Accounts

Abstract

This project aims to assist brands and influencers in categorizing Instagram accounts as either "Active" or "Inactive" based on engagement metrics. Automating this classification process enables businesses to refine their content and engagement strategies more effectively. To build a reliable system, we utilized a dataset containing key engagement metrics, including profile visits, likes, and follows. A synthetic target variable, "Active_Status," was created by establishing specific thresholds for these metrics, facilitating user activity classification. For analysis, we employed three widely used machine learning models: Logistic Regression, Support Vector Machine (SVM), and Gradient Boosting Classifier, selected for their effectiveness in classification tasks. The dataset was split into 70% for training and 30% for testing, with data scaling performed using Standard Scaler to ensure uniform feature treatment. After training, model performance was assessed using accuracy, precision, and confusion matrices to determine their effectiveness. Additionally, visual tools such as charts and graphs were incorporated to enhance result interpretation. Among the models tested, the Gradient Boosting Classifier demonstrated superior performance due to its ability to sequentially construct multiple decision trees and refine its predictions by learning from errors. This capability allowed it to detect subtle engagement patterns that simpler models like Logistic Regression and SVM might overlook. Given its robust classification accuracy, the Gradient Boosting Classifier proved to be the most reliable model for distinguishing active and inactive Instagram users, providing valuable insights for businesses and influencers to optimize their social media strategies.