Car Evaluation

Evaluate a car based on its characteristics using classification with Tensorflow2.0.

Background

Tensorflow2.0 is the latest version of Google's flagship deep learning platform.

• Uses Keras API as its default library for training classification and regression models.
• Problem with earlier versions of TensorFlow was the complexity of model creation.

Dependencies

• TensorFlow 2.0
• Pandas
• Numpy
• Matplotlib
• Seaborn
• Scikit-learn

pip install -r requirements.txt

To make sure TensorFlow2.0 is installed: pip install --upgrade tensorflow

Dataset

• Car Evaluation Dataset: https://www.kaggle.com/elikplim/car-evaluation-data-set
  • Saved in: data/car_evaluation.csv
• Details: https://archive.ics.uci.edu/ml/datasets/car+evaluation
  • price: the buying price of the car)
  • maint: the maintenance cost
  • doors: number of doors
  • persons: the seating capacity
  • lug_capacity: the luggage capacity
  • safety: measure of how safe the car is
  • output: the condition of the car

Data Analysis

Distribution of output

• 70% are in unacceptable condition
• 20% cars are in acceptable conditions
• The percentage of cars in good and very good condition is very low

Distribution of values of other features

• Price: low, medium, high, very high: each 25%
• Maintenance: low, medium, high, very high: each 25%
• Doors: 2, 3, 4, 5 or more: each 25%
• Persons: 2, 4, more: each 33.33%
• Luggage capacity: small, big, medium: each 33.33%
• Safety: low, medium, high: each 33.33%

Data Preprocessing

All the features are categorical

• Convert the categorical columns into numeric
  • One-hot encoding: For each unique value in the categorical column, a new column is created. For the rows in the actual column where the unique value existed, a 1 is added to the corresponding row of the column created for that particular value.
• Prepare features and labels

Split data into Training and Test sets

• Training set = 80%
• Test set = 20%

Create model

• Using Keras functional API
• Input layer
• 2 Hidden layers with 15 and 10 neurons respoectively and ReLU activation function
• Output with 4 neurons for 4 label values and Softmax activation function
• Loss function = categorical cross-entropy
• Optimizer = Adam
• Evaluation metric = accuracy

Train the model

• Epochs = 100
• Validation data = 20% of training data

Evaluate the model

• Accuracy score
• Loss