Skip to content

lequibao/driving_behavior

 
 

Repository files navigation

driving_behavior

This is under review.

AI tachograph is able to detect abnormal driving behavior with over 0.9999 prediction accuracy on average in real time. Data capturing system is capable of collecting abnormality data from 10181 commercial vehicles and 12530 drivers in real time in 0.1 second increments, 24 hours a day, 365 days a year. We have recently added a new normal data capturing function to the data capturing system for machine learning. Normal data can be captured and collected every minute from 9169 commercial vehicles and 6301 drivers.

Python source programs

rfcv.py with 493 trees and entropy criterion is a Python program for 10-fold cross-validation using 24_27rand.csv.

rfcvgini.py with 493 and Gini impurity criterion is a Python program for 10-fold cross-validation using 24_27rand.csv.

evalcross.py is a 10-fold cross-validation Python program that runs rfcross.py 100 times, randomly generating the number of trees ranging from 150 to 500.

rfcross.py is a Python program that specifies the dataset file and the number of trees in a in random forest binary classification.

datasets

24_27rand.csv can be used for this experiment. It will be available on the request.

results

Entropy: 10-fold cross-validation with random forest binary classification of 493 trees and entropy criterion was: [0.99993464 0.99993464 0.99991285 0.99986928 0.99993464 0.99989107 0.99995643 0.99993464 0.99989107 0.99989107]. The average accuracy is 0.999915 and the standard deviation is 0.000027.

Gini impurity: 10-fold cross-validation with random forest binary classification of 493 trees and Gini impurity: with Gini impurity was calculated: [0.99993464 0.99986928 0.99989107 0.99984749 0.99993464 0.99989107 0.99995643 0.99993464 0.99982571 0.99984749]. The average accuracy with Gini impurity criterion is 0.999893 and the standard deviation is 0.000043.

evalcross150_500.result

evalcross150_2000.result

About

Artificial intelligence tachograph

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages

  • Python 100.0%