The Infant Mortality Rate (IMR) is the number of infants per 1000 that do not survive until their first birthday. The problem statement was to predict the infant mortality from Pakistan Demographic and Household Survey (2017,2018).The data collected was from Demographic Health Survey (DHS) program organized by USAID. My first step was to upload and extract sample points. I extracted those sample points which were infants. After extracting the infants sample points the data was reduced to 2486 data points with 1186 features. After significant preprocessing and cleaning the data I applied Principal Component Analysis which can be seen below: 

Then I applied different Machine Learning Algorithms which includes Light Gradient Boosting Machine, Support Vector Machine and Logistic Regression

I also applied Deep Learning techniques and trained different models to achieve state of the art results

After training and analyzing, my goal was to make my model as generalized as possible. I used Nepal: Standard DHS, 2016 to test my model on completely unseen data. The data included 378 sample points. After preprocessing, I fit my trained Random Forest algorithm. The model was able to classify all the sample points correctly and gave 100 percent accuracy. This motivated me to deploy the model on web. Finally, after making Flask API, I successfully deployed the model on Heroku for public use.

Link for Model: https://infantmortalityapi.herokuapp.com/