Apartment Price Prediction of Bangladesh Using Machine Learning and Deep Learning Algorithms

Abstract

The overall aim of this study is to analyze the effectiveness of different machine learning and deep learning models for apartment price prediction in terms of their accuracy and error measures, so as to determine how well these models can predict apartment prices. The experimental setup was done using exhaustive preprocessing and utilized the Scikit-Learn, TensorFlow frameworks for training the model. Also considered were Random Forest, Decision Tree, K-Nearest Neighbor, Gradient Boosting, Extreme Gradient Boosting and Convolutional Neural Network (CNN), Artificial Neural Network (ANN) among others. They also used evaluation indicators like Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Mean Squared Error (MSE), Accuracy (R²). It was found that the Extreme Gradient Boosting Regressor had the highest accuracy of 97.53%, indicating its reliability in forecasting apartment prices. Following this closely was Decision Tree Regressor with an accuracy rate of 96.68%, proving that it is still efficient even though it is a simpler model. On top of that, Random Forest Regressor has also shown good performance with an accuracy score of 96.15% owing to its ensemble nature. Conversely; reduced prediction accuracies were observed for both K-Nearest Neighbor Regressor and Artificial Neural Network (ANN) having R² values at 87.55% and 89.73% respectively. Despite their lesser accuracy, these models still offered useful insights and could be further improved. The Convolutional Neural Network (CNN) demonstrated competitive performance, suggesting its potentials on intricate patterns in the data. This study points to the best fit of Decision Tree Regressor and Extreme Gradient Boosting Regressor among other models for forecasting apartment prices. These results indicate that selection of a model and preprocessing are important in predicting real estate prices accurately.