next up previous contents
Next: Urban Area Classification Based Up: Computational Information Processing Previous: Compound analysis by ombining

Modelling of Consumer Demand of Stores

Researchers: Tommi Orpana, Jouko Lampinen, Jukka Heikkonen and Laura Sirro


This study belongs the project ``Data-fusion and neural networks in complex models'' financed by TEKES and participating enterprises.

The essence of this study is to determine the flows of money from consumers to stores in a region. By flows we can mean the amount of money consumer A spends at store B, but also more concretely the route which he takes to reach the store. The goal is a statistical model of the long-term average behaviour of consumer groups. The models constructed in this project are not dynamic but rather yearly averages of consumer spending and store sales.

The models are of the multinomial logit type (or variations, depending on the assumed distribution of the nondeterministic component in the consumer utility function). The factors that are assumed to influence the store choise are included in the model and the parameters reflecting the effect of the factors are estimated using statistical methods, so that the model successfully predicts the yearly sales of the stores in the region. We use maximum likelihood and Bayesian methods to determine the parameters and their distributions. The major tasks in model construction are determining the most relevant factors and partitioning of consumers and stores into groups.


 
Figure 6
Figure 6: Figure shows an example analysis of grocery store demand distribution from Kouvola region. Based on store locations (black dots) and population distribution (colored house icons) and total sales of the stores, the developed system estimates the utility parameters for different store types. In the figure the red box with white asterisk shows the largest hypermarket in the region, and the colors of the house icons indicate the predicted share of money spent in that hypermarket of all the funds used in grocery stores. The width of the map area is about 50 km.


next up previous contents
Next: Urban Area Classification Based Up: Computational Information Processing Previous: Compound analysis by ombining
www@lce.hut.fi