M3lvin005/PulseIQ-Africa

# PulseIQ Africa **Decision-intelligence web app for small businesses and financial teams.** Upload a CSV of business records and get back dashboards, credit-default risk predictions, suspicious-activity flags, and a downloadable PDF report. ## The problem Small businesses and lending institutions in developing economies keep useful data in Excel and messy CSV files, but have no simple way to turn that data into decisions. Commercial BI and credit-scoring tools are priced for enterprises. PulseIQ closes that gap: it takes a spreadsheet as-is and returns performance, risk, and anomaly analysis without configuration. ## What it does **Data quality assessment** — Reports row and column counts, missing values, duplicate rows, and an overall data-quality score before any analysis runs. **Dashboards** — Revenue trends, repayment and default status, risk-level breakdowns, transaction distribution, customer segments, and suspicious-activity categories. **Credit risk prediction** — Trains three classifiers and selects the strongest by F1-score and ROC-AUC. Scores individual customers or loan applications with a decision, a reason, and a suggested action. **Anomaly detection** — Flags suspicious transactions using transparent, inspectable business rules rather than a black-box model, so a reviewer can always see why something was flagged. **Reporting** — Generates a downloadable PDF business-intelligence report via ReportLab. **Insight assistant** — Answers plain-language questions about the dataset using rule-based logic, with no paid AI API required. ## Tech stack | Layer | Choice | |---|---| | Web app | Streamlit | | Language | Python | | Data | pandas, NumPy | | Charts | Plotly | | ML | scikit-learn | | PDF | ReportLab | ## How the model works PulseIQ normalises uploaded CSV headers and assembles a model frame from these fields when present: `income`, `loan_amount`, `repayment_history_score`, `existing_debt`, `transaction_frequency`, `account_age_months`, …