Alphabet Soup has tasked us with buiding a model to provide insight into organizations that are successful in deploying the donations received. Our model will help analyse which organizations are better bets for placing Alphabet Soup's donations. A sizable data set of thousands of organizations has been provided with 12 data points per organization for us to model.

Due to the nature of some of the variables, binning and encoding were required.

"IS_SUCCESSFUL" was provided and was the target variable for our analysis.

Application Type, Affiliation, Classification, Use Case, Organization, Status, Income Amount, Special Considerations and Ask Amount were all retained variables for the modeling.

EID and Name were identification variables that we dropped as they were neither targets or features.

Eventually, many combinations of neurons, layers and activation functions were modeled, but the most successful models all topped out around 72%.
-In the end, three hidden layers with 80, 60 and 30 nodes respectively are shown in the submitted model.
-Activations for the hidden layers were relu functions, with a sigmoid used in the output layer.
-Functionally, in working toward the model target of 75% accuracy, we added a hidden layer and increased neurons, but at a marginal gain over the initial challenge guidance. Removal of certain feature variables was also attempted, but with insufficient success to meet the model target.

Unfortunately, the iterations and adjustments to the model were unsuccessful in my attempts to meet the target model accuracy of 75%. As this was a binary modeling exercise in identifying high risk organizations, a model distinguishing into 2 classes, separating high risk from more effective charities, could lead to a better outcome given the task at hand.