umerijadon/HealthConnect-Project

Binary classification project predicting patient no-shows for HealthConnect Clinic using historical appointment and demographic data — built during the AnalystLab Africa Data Science track. # HealthConnect-Project --- ## 📌 Problem Overview & Business Context HealthConnect Clinic faces operational inefficiencies and financial losses caused by patients missing scheduled medical appointments. Unattended appointment slots result in: 1. **Wasted Operational Capacity:** Unused medical staff hours and clinical facilities. 2. **Delayed Patient Care:** No-shows prevent other patients on waiting lists from accessing timely healthcare. 3. **Revenue Leakage:** Unutilized time slots directly impact clinic sustainability. ### 🎯 Central Project Question *Can HealthConnect Clinic leverage historical appointment and patient demographic data to predict, at or before booking time, whether a patient will fail to attend a scheduled appointment?* By formulating this business problem as a **binary classification model**, the clinic can identify high-risk appointments in advance and deploy targeted intervention strategies (e.g., automated SMS/WhatsApp reminders, direct telephone follow-ups, or strategic slot overbooking). --- ## 📊 Dataset & Exploratory Data Assessment The dataset (`HealthConnect_Appointment_Data.csv`) contains **5,000 anonymized historical records** spanning 18 distinct variables. ### 1. Data Structure & Summary Statistics | Metric / Feature | Value / Details | | :--- | :--- | | **Total Record Count** | 5,000 records | | **Unique Patients** | 1,696 patients | | **Feature Count** | 18 columns (Numeric, Categorical, Datetime) | | **Primary Key / ID Integrity** | `appointment_id` (0 duplicates) | | **Target Variable Distribution** (`appointment_outcome`) | **No-Show:** 48.5% (2,423)  •  **Attended:** 46.3% (2,314)  •  **Cancelled:** 5.3% (263) | ### 2. Data Quality & Missing Value Assessment * **`reminder_channel`** (1,366 missing / 27.3%): Structural gap. Cross-tabulation confirms missingness occurs *exclusively* when `reminder_sent == 'No'`. Standardized by imputing `'None'`. * **`distance_to_clinic_km`** (90 missing / 1.8%): …