Ambeone Student’s Projects Gallery
Based on the Big Data Analytics & Machine Learning Techniques taught in Ambeone’s Programs
This is a Gallery of some glimpses Data Science projects done by recent Ambeone students as part of their program.In case you are interested to know more about a particular project/projects, you may contact us for details .
Journey Transition Time Analysis & Delay Prediction
Submitted by: Manjusha
Transport Vehicle Transit Time & Delay Prediction Model Using Data Science
Overview
The Transit Time and Delay Prediction Model was developed using data Science Regression Modelling for a leading Transport and Logistics company based in UAE and servicing whole of GCC with their fleet of Trucks. The transport trucks reach their destination ,crossing many checkpoints ( stages) .Time stamp is inserted to the Database at the entry and exit of each journey check point. A standard expected time is assigned to each of these check points and the variance is calculated from the actual and expected time taken by the vehicle to reach the check point .
This data was used to create a Prediction Model using the Time Variances at different check points as the independent variables and the Time variance in reaching final destination as the dependent variable.
Objective
- Predict the transition delay of each journey
- Predict the ratio of delayed trips per month
Data Preparation & Processing
- Handled NAs and Data Entry errors
- Created new variables like day-of-week of the journey
- Calculated the transition time for all stages from entry, exit time
- Divided Dataset into Training & Testing
- Shapiro test was done
- Correlation matrix was plotted
- Created models using different data science methods ,evaluated and improved them
Data Science Techniques and Models used
- Linear Regression
- Time Series
- Neural Network
Key Results
Correlation:
- TT to start journey, TT inside Batha,TT to destination from Batha are well correlated to journey completion time.
- Time taken to Batha from Silla is inversely correlated to the time spent inside Silla.
- Time Taken for Loading Arrival is inversely correlated to time taken for loading.
Linear regression:
Created a model to predict the transition delay for a journey in hours using Linear Regression and Improved the model up to 93.39 R-Square using training and testing data.
Time Series:
ARIMA,ETS,ARIMA-STL,Simple Exponential Smoothening,Holt’s method were applied to predict the ratio of delayed journeys (monthly) and found that ARIMA-STL and ARIMA yielded the best results.
Neural Network:
Further improved the model using NN and RMSE was reduced from 86 to 41.
A Robust Transit Time and Delay model was developed which helped the organisation in accurately predicting the Transit times for various journeys and help manage the predicted delays.
