A Machine Learning Based Movie Status Evaluation System for Bangladesh Movies

Abstract

This study develops a machine learning based on movie status evaluation system for Bangladesh movies. The number of movie production rate is growing day by day worldwide and the movie maker invests highly in the movie industry. In such a scenario, this is very important to evaluate movie status. In our proposed research, we evaluate the status of Dhallywood movie (Bangladesh Cinema) based on three different types of Machine Learning (ML) based classification, Binary classifier that includes two targeted classes, Triple classifier that includes three targeted classes and, four classifier that includes four targeted classes. Here, we will give our detailed analysis of data because Bangladeshi movie data collection is the main challenge of our work and consequently, we analyze our data in different ways to set target variable to improve the accuracy of models. For the first time any research focuses on Dhallywood movie data where we have used different machine learning based models for analyzing data. We apply the same ML algorithm for each of the three different class classifications to find which classifier is performing well for our data and the problem by comparing the obtained accuracies. From the experiments, it is observed that the triple class classification accuracy is higher than binary and four class classifications. Among the five applied ML algorithms, the Random Forest shows the best accuracy around 85%. Our research provides a quite different approach to set target variable class based on Wikipedia data, news, actor- actress biography, and viewer response on YouTube for a particular movie. We go for this approach because Bangladeshi movie rating is not perfect on IMDb also the budgets and revenues are not found for all movies.