Aim

$$ Apply concepts defined in class on preprocessing, data management and classification.

$$ Implement and configure artificial neural networks $($ANN$)$ in Python

$$ Analyze and synthesize the results

Description

Implement artificial neural networks to classify a set of data assigned by the teacher. The project must do the following:

The following neural networks will be used:

$1.$ MLP $($Multi$-$Layer Perceptron$)$

$2.$ CNN $($Convolutional Neural Network$)$

$3.$ Hybrid: combination MLP$,$ CNN$,$ RNN $($Recursive Neural Network$)$

Class balancing will be done using the Borderline$-$Smote method.

Methodology

$1.$ Part$-1$ Preprocessing:

For each assigned neural network $($MLP$,$ Hybrid, CNN$),$ use data from $1$ patient and do the following:

$1\mathrm{.}1.$ Perform classification using $$Raw data$,$ class balancing using the Borderline Smote method and CV$=6$

$1\mathrm{.}2.$ Perform classification with standardized data $($StandardScaler$),$ Normalized, class balancing and CV $=6$

$1\mathrm{.}3.$ Compare the results between $1\mathrm{.}1$ and $1\mathrm{.}2$ using only the accuracy metric. Choose the best method and answer the questions in the report $($Part: Results item $1\mathrm{.}2)$

$1\mathrm{.}4.$ Get the best Hyper$-$parameters $($learning rate, bactch$\_$size, epoch, etc..$)$ by doing several tests or using GridSearch to get the highest accuracy.