In our Logistic Regression our goal was to see if the inputs could produce the correct "Label" of having, or not having Autism. After running the first model we realized that ten of the inputs, which were survey questions given to users, directly affected the classification of Autism, or not Autism. Our accuracy was nearly 100 % which gave us pause. We ranked the feature importances of the data and confirmed that the survey questions are the most influential in predicting the class. Because this skewed our model we ran two more models. One model took in 5 of the survey questions. The other model looked at gender and whether there was a family history of Autism. An interesting finding we made was that using only 5 of the survey questions, the model returned a testing score with 90 % accuracy. This could have implications if a shorter survey were to be designed. The model with the inputs of gender and family history of autism gave us a 68 % accuracy. For each of these models, we also created a Confusion Matrix, which details how many times Autism was predicted successfully, with a True Positive, or misdiagnosed with a False Positive. It also lists the same information for a True negative, and False-negative when making a diagnosis of autism. Please see the Jupyter Notebooks for our code, plots and markdown notes!