codevshl / iheart_optum_stratethon_s4

iHEART_OPTUM_STRATETHON_s4

FOR FULL DEMO WITH MY ANOTHER SIMILAR PROJECT

Problem Statements

• Pricing transparency and bring affordability in cost of care for long-term care patients (Chronic care diseases).

• Reducing hospitalization (and re-hospitalization) risk through the early intervention program

Problems Solved in Multiple Stages

First phase of webapp rollout

• Lifestyle based risk assessment .

  • This prevents possibilities of hospitalisation with early intervention with user friendly interface anytime anywhere solving our first problem statement.
  • The user needs to answer a detailed questionnaire about diet, lifestyle, medical history.
  • No dependence on the results of the blood test or any clinical data.

• Hospitalization (risks of unwanted incidents) [this too is a type of early interventions program] .

  • Cross verifies the possible heart disease in first step.
  • Can be monetised for making prediction if user isn't using our organisation's services.
  • Provide member user free prediction of possible heart disease and premium feature of classification of heart disease using our model.

Second phase of webapp rollout

• Diagnostic (risks of incorrect diagnostics) .

  • Provides various treatment packages considering competitors offers too and taking long term care of our patient for chronic disease which will solve our second problem statement.
  • Prevents possibilty of early discharge and eventually of re-admission/hospitalisation solving our first problem statement.

• Rehabilitation (risks of rehabilitation defects) & provide proper care plan .

  • Devising appropriate patient-care plans Provides care plans for Rehabilation which will solve second problem statement.

• Customer Feedback (to keep check on services provided and improvements required).

Third phase of webapp rollout

• Re-hospitilazed patients reason finding with the tracked data using IOT and patient history.

• Medical (risks of surgical treatment, risks of pharmacotherapy, risks of undesirable medication reactions) .

CODE FLOW FOR PREDICTION MODEL

1. Add data using CSV file.
2. Data preprocessing.
3. Data Cleaning and Normalising.
4. Data Visualisations using different Data Plot types.
5. Split Data in two parts. 
  - Training Data.
  - Testing Data.
6. Train model.
7. Evaluate the model using following Algorithms:
  - Random Forest.
  - KNN.
  - SVC.
  - Logistic Regression.
    > Found Random Forest most accurate with Accuracy = .99, Recall = .98, F1 Score = .99, Precision = 1.00. 
8. Read CSV file or Manual Data from user.
9. Predict using our Model.  ```