Download presentation
Presentation is loading. Please wait.
Published byUtami Kartawijaya Modified over 6 years ago
1
NEURAL NETWORK APPROACHES FOR AUTOMOBILE MPG PREDICTION
ECE 539 Course Project NEURAL NETWORK APPROACHES FOR AUTOMOBILE MPG PREDICTION 12/14/2010 Xiaofei Sun University of Wisconsin-Madison
2
Motivations Nowadays, fuel economy becomes a great concern of the governments and drivers MPG varies with vehicle specs and conditions Database available online only accounts for different models Large amount of data required Build NN models to predict the MPG based on given specs and conditions MLP RBF 1/8 1
3
Data Description 1 Output: MPG Source: UCI Machine Learning Repository
8 Inputs: 1. cylinder # 2. displacement 3. horsepower 4. weight 5. acceleration 6. year 7. origin 8. manufacturer 1 Output: MPG
4
Data Preparation 392 sets of data
Correlation coefficients between I/O were calculated
5
Linear Regression 7-way cross validation Training MSE = 11.12
Tuning MSE = 12.70
6
Multi Layer Perceptron
MATLAB Neural Network Toolbox Used Learning algorithms: Gradient descent with momentum Scaled conjugate gradient Levenberg-Marquardt Datasets were randomly divided into three subsets: 60% for training 20% for validation (early stopping) 20% for testing
7
Multi Layer Perceptron
Structure: feedforward network Log-sigmoid function for hidden layer Linear function for output layer Test MSE = 5.11 Training MSE = 4.03
8
Conclusions and Future Work
MLP yields better performance than linear regression after fine tuning Will construct radial basis function network, and compare with MLP
9
? Any Questions?
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.