Research on Commercial Vehicle Replacement Prediction Model Based on Improved Attention Mechanism

As the automotive industry has experienced rapid development, vehicle replacement has become an important issue for automakers and dealers. Accurately predicting customer replacement demands is of great significance for automakers and dealers to formulate effective marketing strategies. This paper proposes an improved attention-based commercial vehicle replacement prediction model. First, it constructs the temporal and non-temporal features of vehicle operations, then uses bidirectional LSTM and MLP networks to extract high-dimensional features, and utilizes an improved attention mechanism to concatenate the high-dimensional temporal and non-temporal features. Finally, it outputs the vehicle replacement prediction results through a fully connected layer. The experimental results show that this method can predict vehicle replacement results for the next three months with a weighted accuracy of 94.43%, providing data support for proactive marketing by automotive sales companies.