Comparison of LSTM Model Performance with Classical Regression in Predicting Gaming Laptop Prices in Indonesia

Abstract

The demand for gaming laptops has surged in the digital era, appealing to both professional gamers and the general public. Gaming laptops come equipped with advanced features such as powerful graphics, fast processors, and sleek designs, offering a portable solution for gaming enthusiasts. However, the price of gaming laptops varies due to factors like brand, hardware specifications, screen size, and additional features. Accurately predicting these prices can help consumers make informed purchasing decisions and assist manufacturers in setting competitive prices. This research proposes the use of the Long Short-Term Memory (LSTM) algorithm to predict gaming laptop prices, comparing its performance with classic regression algorithms such as Linear Regression and Multi-layer Perceptron. Utilizing a comprehensive dataset of gaming laptop prices and specifications in Indonesia, this study employs robust pre-processing and model optimization techniques. The results show that the LSTM model achieves a Root Mean Squared Error (RMSE) of 0.09011, a Mean Squared Error (MSE) of 0.00812, and an R² Score of 0.90016. In comparison, the Linear Regression model has an RMSE of 0.09075, an MSE of 0.00823, and an R² Score of 0.89873, while the Multi-layer Perceptron model has an RMSE of 0.09891, an MSE of 0.00978, and an R² Score of 0.87971. These results indicate that the Long Short-Term Memory algorithm outperforms other classic regression algorithms in this case. This study highlights the potential of LSTM in developing a robust price prediction model for gaming laptops, particularly in the Indonesian market, providing valuable insights for both consumers and manufacturers.

Keywords

• classical regression
• gaming laptops
• lstm
• price prediction
• regression