Market segmentation through data mining: A method to extract behaviors from a noisy data set

- A method to segment markets based on delivery data is proposed.
- Similarities between customers are calculated with Dynamic Time Warping.
- Segments reveal behavior patterns not recognizable in the very noisy original data.
- The behavior patterns are useful for developing long-range forecasts.

Strategic business planning requires forecasted information that contains a sufficient level of detail that reflects trends, seasonality, and changes while also minimizing the level of effort needed to develop and assess the forecasted information. The balance of information is most often achieved by grouping the customer population into segments; planning is then based on segments instead of individuals.Ideally, separating customers into segments uses descriptive variables to identify similar behavior expectations. In some domains, however, descriptive variables are not available or are not adequate for distinguishing differences and similarities between customers. The authors solved this problem by applying data mining methods to identify behavior patterns in historical noisy delivery data. The revealed behavior patterns and subsequent market segmentation are suitable for strategic decision-making. The proposed segmentation method demonstrates improved performance over traditional methods when tested on synthetic and real-world data sets.