| Summary: | The oil and energy sector is a very traditional, controversial and competitive sector. This study is based on a Portuguese fuel company and its main objective is to identify and characterize potential variables with predictive capacity for sales of new fuel stations. The database consists of a set of context variables with predictive potential for sales of fuel stations and monthly sales in terms of fuel volume. The research methodology focused on statistical methods of exploratory data analysis, clusters analysis and regression models. The fuel station context variables tend to characterize the socio-economic conditions of the area of influence of each station, such as population density variable, others related to the similar existing supply of both the company itself and the competing companies, and others related to geographical location and accessibility. The exploratory data analysis allowed to identify several patterns in the time series of sales indicating that the investigation of factors must be segmented. Homogeneous groups of fuel stations were identified through a hierarchical agglomerative clustering procedure considering the Ward's minimum variance method and the square Euclidian distance as distance measure. For each of the groups identified, multiple linear regression models were adjusted considering the annual fuel sales in the 1st, 2nd and 3rd years of operation of the stations as dependent variables. The results show that not all the exogenous variables are statistically significant. However, it is possible to conclude that the average daily traffic is the variable with predicted capacity for the most of the groups of fuel stations analyzed.
|
|---|