Doctor Prescription Service

Doctor Prescription Service is a machine learning-based project designed to classify diseases based on a given set of symptoms and provide corresponding prescriptions. The aim is to create an intelligent system that assists doctors and patients by offering accurate diagnoses and treatment suggestions.

Table of Contents

• Project Overview
• Features
• Installation
• Preview
• Data
• Model Training
• Prescription Generation
• Contribution
• License
• Contact

Project Overview

In this project, we develop a machine learning model to classify diseases based on symptoms provided by the user. Initially, a disease will be predicted, but later on, we are working to also provide prescriptions for the same (possibly using another model).

Features

• Symptom-based Disease Classification: Classifies diseases based on symptoms provided by the user.
• Prescription Generation: Provides a list of medications and treatment plans tailored to the diagnosed disease. (tentative)
• Simple Interface: A simple web interface will be used to allow easy and controlled interaction with the model.
• User Login: User login along with chat history is also provided.

Installation

To properly ensure that the project runs properly, clone the repository and contact me for the secret api keys, contact details are given below. Use commands like npm start after ensuring all packages have been installed.

Preview

Signup Page:

Interface After logging in:

Prediction:

Chat History:

Data

The project relies on a dataset containing symptoms and their associated diseases. The dataset has the following:

• Symptoms: A list of columns containing symptoms associated with each disease. Each column is a symptom on its own with a binary output: 0 representing that the symptom is not present, and 1 meaning that the symptom is present.
• Diseases: A comprehensive list of diseases covered by the model.
• Prescriptions: A list of drugs that can be used to help in the treatment process. (tentative)

The dataset has around 250,000 rows comprising 339 symptoms with 773 diseases.

Model Training

The machine learning model is trained using the following steps:

1. Data Preprocessing:

   • Clean and preprocess the symptom data.
   • Encode the symptoms and diseases as numerical values.

2. Model Selection:

   • Currently RandomForestClassifier came up as the best suited algorithmn for our use case however we also use:
     • Logistic Regression
     • Boosting techniques (like Gradient Descent, Histogram Boosting)

3. Training:

   • Train the selected model on the dataset.

4. Evaluation:

   • Evaluate the model's accuracy, precision, recall, and other metrics. A comprehensive graph will be provided in the Jupyter notebook.

5. Optimization:

   • Optimize hyperparameters using techniques like Grid Search or Randomized Search.

Prescription Generation

Once a disease is classified, the system generates a prescription (method not decided yet).

Contribution

This project is jointly contributed to by the members of the DPS Team in Makernova 2.0, a recruitment program for DRISHTI (SVNIT).

License

This project is licensed under the MIT License.

Contact

For any inquiries or further information, please reach out via email:

Hetvi Shah: [email protected]

Feel free to connect for collaboration, project queries, or access to relevant resources such as API keys.