Sales forecasting of short-life cycle products using clustering and classification techniques

Technology and globalization have created strong competition among companies in many industries. This has led to the rapid development of technological products resulted in a shorter life cycle. Forecasting is essential to decrease supply chain costs; however, traditional methods of forecasting are not suitable in this situation because of the lack of historical data and volatile demand. The literature addresses this challenge using two stages of data mining techniques by clustering the existing products then classifying the new ones into these clusters. This paper adopts this technique by using the K-means algorithm to cluster products with similar sales profiles. Then, the rule induction is used to generate the criteria of the product in each cluster. The generated rules are used to forecast the sales profile of a new product using the assigned clusters. A case study is used to validate the algorithm by comparing the proposed technique with the literature shows significant improvement by using the suggested preprocessing and rule induction.