ML and Ai PCOS Risk health detection

PCOS Risk Detection – Implementation Plan for Our Memmzy Health App Step 1: Define Problem & Scope (4–5 hours) Objective: Predict PCOS risk based on cycle, symptom, and hormonal data. App Use Case: Users enter their cycle length, irregularities, symptoms (acne, hair growth, weight gain), lifestyle, and optional lab values (FSH, LH, AMH, insulin, testosterone) . The ML model provides “Low / Medium / High risk for PCOS” with educational guidance. Output: Model is screening / risk scoring tool (not diagnostic). Step 2: Collect & Explore Data (8–10 hours) Datasets: PCOS Dataset on Kaggle (~541 records with clinical + cycle data) Supplement with synthetic data generation (to expand features and simulate app data). Work: Data cleaning (handle missing values, normalize ranges) Exploratory Data Analysis (EDA): correlations between features and PCOS presence Deliverable: Clean dataset in .csv hosted in GitHub repo Step 3: Feature Engineering (6–8 hours) Inputs: Demographics: Age, BMI Cycle Data: Cycle length, irregular periods, missed cycles Symptoms: Acne, hair growth, weight issues, infertility history Lab Values (optional): FSH, LH, AMH, Testosterone, Insulin Processing: Normalize continuous values (z-score / min-max scaling) Encode categorical features (One-Hot Encoding) Handle missing lab results (imputation strategies) Deliverable: Preprocessed dataset + scripts Step 4: Model Selection & Training (10–12 hours) Baseline Models: Logistic Regression, Random Forest, Gradient Boosting (XGBoost/LightGBM) Advanced: Neural network (if dataset expanded) Steps: Train/test split (70/30) Cross-validation (k-fold) Evaluate with AUC, F1-score, Recall (important for medical screening) Output: Trained PCOS risk model saved as .pkl Step 5: Model Deployment API (8–10 hours) Tech: Python + FastAPI/Flask Workflow: User submits cycle/symptom data via the Memmzy app API processes and sends to ML model Model returns PCOS risk category (Low/Medium/High) App displays personalized insights + educational content Deliverable: REST API (Dockerized) Step 6: Integration with Memmzy App (6–8 hours) Frontend: Form to collect cycle, symptoms, optional labs Risk output visualization: gauge meter / color-coded risk levels Backend: Connect app backend → PCOS API endpoint Store anonymized data for retraining (compliant with GDPR/HIPAA) Deliverable: Working feature in app test environment Step 7: Testing & Validation (6–8 hours) Unit Tests: Input validation, missing values, incorrect formats Model Tests: Compare predicted vs actual PCOS outcomes in validation dataset App Tests: End-to-end flow (user → API → insights) Deliverable: Test suite in GitHub + CI/CD setup Step 8: Deployment & Documentation (4–5 hours) GitHub Workflow: Branching strategy: dev → staging → main CI/CD with GitHub Actions for automated builds/tests Docker image for API deployment (can run on cloud: AWS/GCP/Azure) Docs: README.md with setup instructions API documentation (Swagger/OpenAPI) Deliverable: Public/Private GitHub repo with clean code + docs Total = ~58–66 hours This plan ensures the PCOS feature fits into Memmzy’s existing AI-driven health ecosystem , while being lightweight enough to implement within ~60 hours.

Digital Content Creation For Health App Geared Towards Women's health

Our company currently don't advertise. We hope to revamp our social media and email marketing content to attract more users to explore the new feature we are launching soon. We would like students to help us create content that is aligned with our vision, mission, and sector. This will involve several different steps for the students, including: Familiarizing themselves with our App, marketing goals and target market. Researching factors affecting the quality of digital content. Seeking and understanding current trends. Recommending changes to existing digital content and designing new social media posts and emails. Recommending and designing new digital content not already used by our company. Bonus steps in the process would also include: Testing and improving designs based on feedback.