This is a simple data cleansing tool that parses information from a hospital's CSV file and converts it into useful information for machine learning algorithms aimed to predict the probability of a patient dying.

First, the tool removes unnecessary columns that should not effect the outcome of the prediction. These columns include (but are not limited to):

• Cost associated with administrator time (dollars)
• Size of dose of medicine, since this data was the same for every patient
• Education and income of patient
• Day of announcement of important updates

Then, the tool reforms some columns, such as the sex column which has mixed data (male, Male, M, female, and 1).

The tool also judges values that are out of the standard expected range, such as a heart rate in the thousands, and removes invalid data.

It also parses categorical data into numeric formats. This data comes from the R^2 test performed on the specific data against the death of a patient.

Finally, the tool also performs statistical analysis and sees if specific columns have a correlation with the death of a patient. This step is done manually and the specific columns must be analyzed by a human to see if they are relevant to the prediction. Columns with bad correlation can be removed in the first step mentioned above.

For example, when faced with data that a wide range, we decided to use a hyperbolic tangent function with some transformations in order to accurately provide our machine learning model with precise data. We also utilized R squared values in order to predict correlation between datapoints and effectively remove them in order to offer better data to our model.

This tool is written in Python and requires the following libraries:

• pandas
• numpy
• matplotlib
• scikit-learn