Neural_Network_Charity_Analysis

Overview of the project:

In this project, we use Deep Machine Learning and Neural Networks to create a binary classifier that is capable of predicting whether applicants will be successful if funded by Alphabet Soup.

From Alphabet Soup’s business team, we 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, such as the following:

• EIN and NAME—Identification columns
• APPLICATION_TYPE—Alphabet Soup application type
• AFFILIATION—Affiliated sector of industry
• CLASSIFICATION—Government organization classification
• USE_CASE—Use case for funding
• ORGANIZATION—Organization type
• STATUS—Active status
• INCOME_AMT—Income classification
• SPECIAL_CONSIDERATIONS—Special consideration for application
• ASK_AMT—Funding amount requested
• IS_SUCCESSFUL—Was the money used effectively

This project consists of three technical analysis deliverables to analyze and classify the success of charitable donations:

• Preprocessing Data for a Neural Network Model
• Compile, Train, and Evaluate the Model
• Optimize the Model

Results :

Data Preprocessing :

• The target variable for my model is the "Is_Successful" column which tells us whether the money was used effectively.

• The features variables in this model are every column for the target, the name and EID columns:
  • APPLICATION_TYPE
  • AFFILIATION
  • CLASSIFICATION
  • USE_CASE
  • ORGANIZATION
  • STATUS
  • INCOME_AMT
  • SPECIAL_CONSIDERATIONS
  • ASK_AMT

• The name and EID columns are neither targets nor features, and should be removed from the input data.

Compiling, Training, and Evaluating the Mode :

• I started with 2 hidden layers and then extended it to 4 hidden layers with various number of Neurons for each layer and tested with different activation functions such as 'relu' and 'tanh'. None of the attempts made a significant difference in the end results for Model Accuracy.

Model	Accuracy	Layers	Hidden Layer1 Neurons	Activation Function1	Hidden Layer2 Neurons	Activation Function2	Hidden Layer3 Neurons	Activation Function3	Hidden Layer4 Neurons	Activation Function4
Model1	74%	2	80	relu	30	relu	-	-	-	-
Model2	74%	3	80	relu	30	relu	20	relu	-	-
Model3	74%	3	80	relu	30	tanh	20	tanh	-	-
Model4	74%	3	80	relu	40	tanh	20	tanh	10	tanh

• First Model

• attempt1

• attempt2

• attempt3

• I was not successful in reaching out the target model performance within my 3 attempts,these attempts did not improve the accuracy results which were around 74%.

• In my first attempt to increase model performance I looked at Income_Amt value counts for binning and determined which values to replace if counts are less than 500. In my second attempt I added a 3rd layer with different number of neruons to each layer. The next few steps I took was to add extra layer and tested different activation functions (Relu and Tanh) to achieve an accuracy higher than 75% .

Summary :

In this project, the goal was to create a model to predict the success of candidate companies receiving donations with use of a deep learning neural network model. We couldn’t get a satisfying performance to help predict the outcome of the charity donations. To increase the performance of the model, we applied bucketing to the Income_Amt values. We also increased the number of hidden layers and their neurons. We played with the activation functions and tested different combinations of RELU, Tanh and Sigmoid. Since we are in a binary classification situation, the recommendation is to add more data and use logistic regression or even random forest classifiers to to generate a classified output and reach our goal.