CNN_Pima_indian_classification

Classify Pima indian dataset with Keras (testing for comparing extracting information with informal method using CNN as tested)

pre-processing

• mean impution on each category
• BaggedLOF
• SMOTE

Model-pre-process_input_output

CNN - Deep learning needed input into format as (instance, row, col, channels)

• padding (length = 9)
• reshape (max_lenght, 3, 3, 1)
• transform output into category.

Model-CNN

• first layer : input_size = (3,3,1)
• second layer : Convolunation 2 dimension with 64 filters and 2 kernel
• thrid layer : Convolunation 2 dimension with 32 filters and 2 kernel
• fourth layer : Max pool size = 1
• Flatten
• create 2 hidden layer with 100 nodes (fully connected)
• output 2

The model is interesting in the coverage of learning rate.

Need to be improve.