Enhancing Automotive Spare Parts Demand Forecasting: A Synergistic Approach Using K-Means Clustering and LSTM-Attention Models

This study aims to enhance demand forecasting in the automotive spare parts industry by integrating K-means clustering with the LSTM-Attention model. Initially, the research applied K-means clustering to segment data effectively, followed by employing the LSTM-Attention model to forecast demand, benchmarking its performance against conventional models. The results indicated significant progress in precision of prediction. While K-means clustering silhouette score of 0.621 indicated that it was successful in sorting out information; the LSTM-Attention model showed outstandingly good results with an MSE 0.052 as well as an MAE 0.182 which are much better than what can be achieved by Simple Recurrent Neural Networks or traditional LSTM methods, confirming their reliability in forecasting demand, especially during periods when there are recurring patterns. With both methods combined, the resulting forecasts were way more accurate than before. It is key to note that all these steps mean a lot to auto parts manufacturing because they give us accurate predictions and control over stock levels. By using this technique one could optimize inventory thereby increasing the total output while minimizing wastage due to unnecessary stock-outs.