For this project, I collected real life dataset about people's preference on various product category using Google Form and applied Exploratory Data Analysis to gain useful insight from the data and see different trends related to Age, Gender, State, Education Level and Employement Status. I also used Mutual Information as part of feature selection to figure on which features the product category depended.
After performing this data analysis, I used one-hot encoding to convert categorical values into numerical useful into my neural network. The neural network consisted of various dense layer with "relu" activation function and using "softmax" activation function in the final layer. I used "adam" optimizer along with "sparse_categorical_crossentropy" as the loss function to achieve the highest accuracy of about 0.64. I also tried age_binning to further improve the accuracy of the model.
After this I used SHAP values to figure out upon which features did the product category depended upon according to the model.