Relating product sales with product purchase arrival times using a Non-Homogeneous Poisson ProcessTools Papadaki, Kalomoira (2020) Relating product sales with product purchase arrival times using a Non-Homogeneous Poisson Process. [Dissertation (University of Nottingham only)]
AbstractMarket segmentation lies at the heart of marketing and business strategy. There have been strides in the techniques implemented for the purposes of segmentation, however no research has been conducted that investigates segmentation using event series. This dissertation explores whether time segmentation can be conducted on event series using a purchase rate. The characteristics of the resulting segments will be described, the differences will be identified, and key insights will be extracted which could be beneficial for marketing strategy and business goals. The potential application of event series segmentation was researched by fitting a Non-Homogeneous Poisson Process (NHPP) onto product-category event series data extracted from an anonymised supermarket relational database. Fitting the NHPP resulted in the extraction of the purchase rates from the product categories over one specified 15-hour day, which were then segmented using the KMeans clustering algorithm. This resulted in four distinct segments that had different purchase rate behaviours over time. From each segment, the purchase rates of the highest and lowest performing product categories in terms of sales were examined in more detail. Across all segments, there was a peak of purchasing rates at around 15:00, after which the decay in the purchase rate differed across the different segments. The most successful segments in terms of sales and quantity sold (Segments 1 & 2) displayed a slow and steady decrease in purchase rates. This could indicate that successful products show plateaus in the purchase rate and slowly decrease over time instead of sudden fluctuations. Therefore, event segmentation using NHPP can inform business decisions based on customer purchase behaviours. Such applications in business include fine tuning online advertisement timing and product recommendations. This dissertation shows that there is promise in event series segmentation by fitting a NHPP on business transactional data, and further research would add value to business.
Actions (Archive Staff Only)
|
Tools
Tools
![[thumbnail of MSc Business Analytics Dissertation.pdf]](https://eprints.nottingham.ac.uk/style/images/fileicons/application_pdf.png "MSc Business Analytics Dissertation.pdf")
