Neuron

It is a Neural Network Architecture which uses Forward Backward Propagation

Overview

• Given a Data Set, the architecture finds the best learning rate, epoch values and number of hidden layers needed
• It uses Forward Backward Propagation.
• The activation function used is sigmoid.
• Based on the selected parameters, the data is classified and the prediction rate is further analyzed.

Dataset Used

• Car Evaluation : https://archive.ics.uci.edu/ml/datasets/car+evaluation

• Mushroom Data set : https://archive.ics.uci.edu/ml/datasets/mushroom

Neural Structure

This is the neural structure for Car Dataset with 6 inputs, 6 neurons with 1 hidden layer and 4 output. The Prediction Rate was 92.13%

Data Set Preprocessing

• All the numerical attributes were scaled using min max scaler, to have the same range 0-1
• All the categorical values were label encoded and then scaled.

Installation

1. Clone the Repository or Download the Project
2. Navigate to the folder
3. Execute 'python Neuron.py'

Sample Execution

1. Select DataSet

Select DataSet
1. Car DataSet
2. Mushroom Dataset
Select one Dataset from above : 1

2. Find Best Optimal Learning Rate

Finding Best Optimal Learning Rate
For Learning Rate : 0.1 the Prediction Rate is 87.27%
For Learning Rate : 0.2 the Prediction Rate is 89.81%
For Learning Rate : 0.3 the Prediction Rate is 91.90%
For Learning Rate : 0.4 the Prediction Rate is 91.67%
For Learning Rate : 0.5 the Prediction Rate is 91.67%
For Learning Rate : 0.6 the Prediction Rate is 90.28%
For Learning Rate : 0.7 the Prediction Rate is 90.05%
For Learning Rate : 0.8 the Prediction Rate is 90.51%
For Learning Rate : 0.9 the Prediction Rate is 90.51%
Optimal Learning Rate '0.3'

3. Find the Best Epoch Values

Finding Best Epoch Value
For Epoch : 3 the Prediction Rate is 82.64%
For Epoch : 4 the Prediction Rate is 83.33%
For Epoch : 5 the Prediction Rate is 83.56%
For Epoch : 6 the Prediction Rate is 84.49%
For Epoch : 7 the Prediction Rate is 84.95%
For Epoch : 8 the Prediction Rate is 86.34%
For Epoch : 9 the Prediction Rate is 86.57%
For Epoch : 10 the Prediction Rate is 87.50%
For Epoch : 11 the Prediction Rate is 87.96%
For Epoch : 12 the Prediction Rate is 88.66%
For Epoch : 13 the Prediction Rate is 88.66%
For Epoch : 14 the Prediction Rate is 88.66%
For Epoch : 15 the Prediction Rate is 88.89%
For Epoch : 16 the Prediction Rate is 89.35%
For Epoch : 17 the Prediction Rate is 90.74%
For Epoch : 18 the Prediction Rate is 92.13%
For Epoch : 19 the Prediction Rate is 91.90%
Optimal Epoch Value '18'

2. Find Best Optimal Hidden Layers

Finding Optimal no of Hidden layers
For Optimal Layer Count : 1 the Prediction Rate is 92.13%
For Optimal Layer Count : 2 the Prediction Rate is 87.50%
For Optimal Layer Count : 3 the Prediction Rate is 68.06%
For Optimal Layer Count : 4 the Prediction Rate is 68.06%
For Optimal Layer Count : 5 the Prediction Rate is 68.06%
Optimal Hidden Layer Count '1'

Project Struture

ExpMaxML

• Neuron.py - Main Startup File.
• /NeuralArch
  • NeuralNet - Class Implementation of Neural Net architecture

    Charts

• Various Charts

  DataSet

• car.data
• agaricus-lepiota.data