Alphabet Soup nonprofit

A deep learning project using Python, Pandas, sklearn, TensorFlow, Google Colab

Background

The nonprofit foundation Alphabet Soup wants a tool that can help it select the applicants for funding with the best chance of success in their ventures. I will use my knowledge of machine learning and neural networks, using the features in the provided dataset to create a binary classifier that can predict whether applicants will be successful if funded by Alphabet Soup.

From Alphabet Soup’s business team, I have received a CSV containing more than 34,000 organizations that have received funding from Alphabet Soup over the years. Within this dataset are a number of columns that capture metadata about each organization.

Steps Taken

• Data Preprocessing:

  • Our target variable is the IS_SUCCESSFUL column
  • The features for our model are APPLICATION_TYPE, AFFILIATION, CLASSIFICATION, USE_CASE, ORGANIZATION, STATUS, INCOME_AMT, SPECIAL_CONSIDERATIONS, and ASK_AMT
  • The EIN and NAME columns were removed because they are identification variables and not necessary.

• Compiling, Training, and Evaluating

  • Several models were tested, changing the number of hidden layers, nodes, epochs and activation functions.

  • The initial model includes:

    • The first hidden layer with 8 nodes and the activation function relu
    • The second hidden layer with 5 nodes and the activation function relu
    • An ouput layer with the activiation function sigmoid
    • The initial model shown reflects the following:

    

  • The final model includes:

    • The first hidden layer with 7 nodes and the activation function relu
    • The second hidden layer with 5 nodes and the activation function relu
    • An ouput layer with the activiation function sigmoid
    • The final model shown reflects the following:

    

Summary

The models I worked on, before and after optimization, were only able to achieve around 73% accuracy. Several models were tested, changing the number of hidden layers, nodes, epochs and activation functions. Making changes to optimize the model only made slight changes in the accuracy; however, I found that adding too many layers hindered the performance. Unfortunatley, I was not able to reach the target performance. Each model tested would not get an accuracy rate higher than about 73%. I found the first model I started with had the highest accuracy rate. In order to solve this problem, I would need to continue to work on finding the optimal number of nodes and hiddenlayers, as well as the best activation function for each hidden layer.