Estimating heavy trucks’ value at buy-back by analyzing telematic data

This thesis focuses on the data mining process, namely the process of distilling information from raw data. The innovative techniques of data pre-processing, the statistical methods to properly treat and represent the data and the algorithms for supervised or unsupervised problems are described from a theoretical point of view and they are applied to the real business case of obtaining a more precise and objective estimate of truck’ s score at buy back based on the insights gathered from the telematic data. Indeed, CNHi has provided its trucks with two applications, UTP and P&CM, that receive from sensors information relative to internal and external parameters, how the vehicle is driven and if some faults occur. These telematic data are sent to CNHi’s IoT platform, in the specific Microsoft Azure, and from here they can be manipulated. Two formulas are proposed to achieve the goal and they are evaluated and tested on real data, based on a huge number of heavy trucks, used in different European countries, with different year of production and leasing contracts. Then, their relationship is compared through a multiple linear regression and the DBSCAN and the K-means are performed to discover hidden pattern among the variables and their correlation with the formulas.