Mobile Phone Price Prediction Model Based on Deep Learning

Abstract

The task of mobile phone price prediction has high application value. It could help customers make more reasonable purchase decisions and help mobile phone manufacturers to better formulate pricing strategies. The study is focused on the application of deep learning models in smartphone price prediction, and expectation to obtain better robustness and prediction accuracy. Multi-Layer Perceptron (MLP), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU) and Transformer model are used to predict mobile phone prices. Weighted average algorithm is used to implement an ensemble model, in order to compare the performance of the ensemble model and the base model on this problem. According to the experimental results, MLP model and GRU model have lower errors and higher R2 scores, indicating that they perform best in this problem. Transformer model lies second, while overfitting occurs during the experiment. LSTM model performed the worst among all four models. Additionally, the study found that increasing the number of model layers leads to overfitting and a decrease in model performance. This study shows the potential of deep learning models in the task of mobile phone price prediction, and better performance of deep learning models could be expected on larger and more complex datasets.