Predictive analytics of driver turnover

点击翻译按钮获取中文摘要

In the current era of big data, high volumes of a wide variety of valuable data about drivers and routes can be**easily collected at a high rate in the transportation and trucking industry. Embedded in these big data are**valuable information and knowledge. Data science solutions--which apply multidisciplinary techniques such as**data mining, machine learning (including deep learning), and statistical modelling--help discover implicit,**previously unknown and potentially useful information and knowledge from big data. For this Engage Project,**DecisionWorks Consulting Inc has reported that the issue of driver turnover is significant for its clients in the**transportation and trucking industry because it costs roughly $5K to train a new driver. Furthermore, bad**drivers (e.g., those who are hard on equipment have poor safety records or who find driving under anything but**good road conditions difficult) are costing problems such as late or mismanaged deliveries, higher than average**maintenance costs, etc. Hence, having data science solutions to discover knowledge and information about**drivers and routes helps reveal the characteristics of "good" and "bad" drivers on "safe" and "dangerous" routes,**which in turn helps DecisionWorks to enhance its business analytic model for (a) reducing the operating cost of**and (b) enhancing the truck driver environment for its clients. Existing manual or automatic approaches focus**on correlation analysis of predefined parameters that are assumed to be significant in driver safety, behavior**and retention. Unfortunately, such an assumption may not hold in all real-life situations. Other relationships**(e.g., causality) other than correlation may exist among the predefined parameters. Furthermore, there may also**exists some factors that affect driver performance beside than those predefined parameters used in the**correlation analysis. In this proposed work, we plan to design and develop a data solution for predictive**analytics of drivers. In particular, in cooperation with the industrial partners, our solution is expected to**perform pattern matching with a predictive analytic component to forecast good drivers at risk of turnover.