Prediction of delivery times using Telematic data

50 %
50 %
Information about Prediction of delivery times using Telematic data

Published on November 21, 2016

Author: VamshiVennamaneni

Source: slideshare.net

1. Driver Identification and Delivery Time Prediction using Telematics Chaitanya Kartheek M, Sai Rajendra Wuppalapati , Vamshi Vennamaneni Professor: Dr. Kevin Lu Driver Identification • K-means Clustering • Logistic Regression & Support Vector Machines (SVM) Business Case: Commercial fleet owners can increase their operating efficiency by analyzing each driver’s driving behavior and accurately predicting the delivery time using telematics data. In this project, we have tried to • Create driver profiles • Driver identification • Predict delivery time for each driver • Explore factors which affect fuel efficiency Driver Profiles Fuel Efficiency: Special Problems in EE - IoT Spring, 2016 Data and Preprocessing • Data is pulled from the Mojio server by calling the Mojio APIs for trip data and vehicle data • Collected data for around 350 trips logged by 3 different drivers and vehicles (urban and highway driving conditions) • Around 20k rows of event data with several trip attributes like Speed Conclusion • Commercial fleet owners and last mile delivery firms can assign drivers depending on their priorities of quick delivery or cost efficiency • Risk assessment of Usage Based Insurance (UBI) products based on driving profiles • Identification of false insurance claims by using driver identification, driver and vehicle profiles IoT device • Telematics device manufactured by Mojio • It can be plugged into the OBDII port of any car or a commercial vehicle • It has a AT&T sim card along with an accelerometer sensor • Collects and stores the event data in Mojio server in the cloud. Data collected includes but not limited to location, speed, MaxA, MaxD, RPM etc. Driver 1 (Urban) Driver 3 (Urban) Driver 2 (Highway) Avg. Speed 26.3 21.0 50.2 Avg. Fuel Efficiency 13.1 12.25 23.95 Avg. Speed during Heading change 35.97 30.66 34.7 Deceleration above threshold per km 0.373 0.947 0.074 Speed tickets per km 0.059 (65kmph) 0.028(65kmph) 0.005 (120Kmph) Delivery time: . A C T U A L P R E D I C T E D 1 2 3 1 1 1 3 2 0 17 1 3 0 1 73 Model Performance matrix (SVM) K-means Clustering model Special Problems in EE - IoT Spring, 2016

Add a comment