CpGene Favicon Prognosis_Api

Prognosis Analysis API

This API provides functionality for analyzing large CSV files with prognosis data and running machine learning algorithms on selected prognosis values.

Features

  • CSV File Upload: Upload large CSV files containing a 'Prognosis' column
  • Prognosis Analysis: Extract unique prognosis values and basic statistics
  • Machine Learning Algorithms: Run various ML algorithms on selected prognosis values
  • Result Download: Download processed results as CSV files
  • File Management: List, view, and delete uploaded files and results

API Endpoints

1. Upload CSV File

POST /prognosis/upload

Upload a CSV file with a 'Prognosis' column for analysis.

Parameters: - file: CSV file (multipart/form-data)

Response:

{
  "task_id": "uuid",
  "sha1_hash": "file_hash",
  "message": "CSV file uploaded successfully. Analysis started.",
  "filename": "file.csv",
  "file_size": 123456
}

2. Get Prognosis Values

GET /prognosis/prognosis-values/{sha1_hash}

Get unique values from the Prognosis column of an uploaded file.

Response:

{
  "sha1_hash": "file_hash",
  "filename": "file.csv",
  "unique_values": ["High Risk", "Low Risk", "Normal"],
  "total_rows": 1000,
  "prognosis_column_found": true,
  "message": "Found 3 unique prognosis values"
}

3. Run Algorithm

POST /prognosis/run-algorithm

Run a machine learning algorithm on selected prognosis values.

Request Body:

{
  "sha1_hash": "file_hash",
  "selected_prognosis_values": ["High Risk", "Low Risk"],
  "algorithm": "random_forest"
}

Available Algorithms: - logistic_regression - random_forest - svm - neural_network - gradient_boosting

Response:

{
  "task_id": "uuid",
  "sha1_hash": "file_hash",
  "algorithm": "random_forest",
  "selected_values": ["High Risk", "Low Risk"],
  "message": "Algorithm random_forest started for 2 prognosis values"
}

4. Check Task Status

GET /prognosis/status/{task_id}

Get the status of a running task.

Response:

{
  "task_id": "uuid",
  "status": "SUCCESS",
  "result": {
    "sha1_hash": "file_hash",
    "algorithm": "random_forest",
    "accuracy": 0.95,
    "output_filename": "random_forest_High Risk_Low Risk_results.csv"
  }
}

5. List Results

GET /prognosis/results/{sha1_hash}

List all algorithm result files for a given file.

Response:

{
  "sha1_hash": "file_hash",
  "result_count": 1,
  "results": [
    {
      "filename": "random_forest_High Risk_Low Risk_results.csv",
      "file_size": 45000,
      "created_time": 1234567890,
      "download_url": "/prognosis/download/file_hash/filename.csv"
    }
  ]
}

6. Download Result File

GET /prognosis/download/{sha1_hash}/{filename}

Download a result CSV file.

7. List All Files

GET /prognosis/list

List all uploaded CSV files.

8. Remove File

DELETE /prognosis/remove/{sha1_hash}

Remove a file and all its results.

9. Remove All Files

DELETE /prognosis/remove_all?delete_pass=123

Remove all files (requires password).

CSV File Requirements

Your CSV file MUST: 1. Have a column named exactly Prognosis - This is mandatory and non-negotiable 2. The Prognosis column must contain at least one non-null value - Empty/null-only columns are rejected 3. Contain numeric columns for machine learning features - At least one numeric column besides Prognosis 4. Be properly formatted CSV with headers - Standard CSV format with comma separation

⚠️ IMPORTANT: Files without a valid Prognosis column will be immediately rejected during upload with a clear error message.

Example CSV structure:

Age,BMI,BloodPressure,Cholesterol,Glucose,HeartRate,Prognosis
45,25.2,120,200,95,70,Normal
60,30.1,140,250,110,75,High Risk
35,22.5,110,180,90,65,Low Risk

Machine Learning Process

  1. Data Filtering: Only rows with selected prognosis values are used
  2. Feature Selection: All numeric columns (except Prognosis) are used as features
  3. Data Preprocessing: Missing values filled with column means, features scaled
  4. Model Training: 80% of data used for training, 20% for testing
  5. Evaluation: Accuracy and classification report generated
  6. Prediction: Full dataset predictions with confidence scores

Output Files

For each algorithm run, the following files are generated:

  1. Results CSV: {algorithm}_{selected_values}_results.csv
  2. Original data + predictions + confidence scores

  3. Sorted by prognosis values and confidence

  4. Trained Model: {algorithm}_{selected_values}_model.joblib

  5. Serialized scikit-learn model

  6. Scaler: {algorithm}_{selected_values}_scaler.joblib

  7. Feature scaler for consistent preprocessing

  8. Metadata: {algorithm}_{selected_values}_metadata.json

  9. Training statistics, accuracy, feature importance, etc.

Testing

Use the provided test script to test the functionality:

uv run python test_prognosis_analysis.py

This will: 1. Create a sample CSV file with synthetic health data 2. Upload the file 3. Extract prognosis values 4. Run a Random Forest algorithm 5. Display results and accuracy

Example Workflow

  1. Upload CSV: Use the upload endpoint to upload your CSV file
  2. Get Task Status: Monitor the analysis task until complete
  3. Get Prognosis Values: Retrieve the unique prognosis values
  4. Select Values & Algorithm: Choose which prognosis values to analyze and which algorithm to use
  5. Run Algorithm: Start the machine learning task
  6. Monitor Progress: Check task status until completion
  7. Download Results: Get the processed CSV with predictions

Error Handling

The API provides detailed error messages for common issues: - Missing Prognosis column - Invalid CSV format - Empty datasets - No numeric features - Invalid algorithm selection

All errors include helpful messages to guide you in resolving the issue.