Neural Network Charity Analysis

The purpose of this analysis was to create a Neural Network Model to analyze and help predict which organizations would make good investment opportunities.

Project Overview:

1. Use preprocessing to prepare the data for a neural network model
2. Compile, train, and evaluate the model
3. Optimize the model

Resources

• Source of data: charity_data.csv
• Software: tensorflow 2.7.0, scikit-learn 1.0.2, Python 3.7.10, Conda 4.10.3, Jupyter Notebook 6.3.0, Visual Studio Code 1.60.2
• Please see the preprocessing and training code here.
• Please see the optimized code here.

Results

Data Preprocessing

• Target columns: IS_SUCCESSFUL
• Dropped columns: EIN and NAME
• Featured columns: APPLICATION_TYPE, AFFILIATION, CLASSIFICATION, USE_CASE, ORGANIZATION, STATUS, INCOME_AMT, SPECIAL_CONSIDERATIONS, and ASK_AMT

Compiling, Training, and Evaluating the Model

• The first model we ran we started with 2 hidden layers at 80 and 30 neurons respectively, with the activation function of ReLu for the hidden layers and Sigmoid for the output layer.

  • The results of the first model had an accuracy score of 72.6% and a loss of 55.4%.

• After several different attempts to optimize the model, we found the following resulted in the best accuracy score:

  • Created binning for the ASK_AMT column
  • Increased the hidden layers 3 at to 80, 50, and 30 neurons respectively
  • The results produced an accuracy score of 72.6% and a loss of 58.1%. The training accuracy did show a promising accuracy score of 74.6% and a loss of 52.3%.

Summary

After several attempts to optimize the neural network model we did not reach the target accuracy score of 75%. An alternative to this model could be a Random Forest Classifier. The Random Forest models run faster than neural network models and can help avoid the data from being overfitted.