Ikmal Thariq KadafiS.Si.D.

Problem The Poverty Depth Index is typically modeled using global (non-spatial) models that ignore spatial variations.

Solution Implementation of the GWLRS model to accommodate both spatial and non-spatial effects.

Impact The GWLRS model achieves a prediction accuracy of 91% and precision of 88%.

Problem How can we objectively measure digital market share, pricing strategies, and geographic distribution of heavy construction materials in the e-commerce era?

Solution A data-driven analytics dashboard was developed by scraping and structuring marketplace data to map real sales performance, pricing gaps, and geographic distribution of lightweight concrete brands in East Java.

Impact The system reveals actionable market insights, identifying dominant players, pricing strategies, and distribution inefficiencies enabling more targeted and competitive business decisions.

Problem The bank faces a high churn rate of 17.94%, which threatens a total managed balance of $115.4M.

Solution I implemented a predictive model (Random Forest) and a dynamic dashboard to pinpoint critical churn drivers, such as the 23.14x risk increase associated with Bronze loyalty levels.

Impact The analysis successfully isolated $12.42M in balance at risk from 18,000 customers, providing a clear roadmap to protect 11% of the bank's total assets

Problem Despite generating high revenue, the campaign showed low efficiency with a ROAS of 1.27x and significant budget leakage, particularly in underperforming channels like Google Ads.

Solution Analyzed channel performance, audience segmentation, and campaign effectiveness to recommend budget reallocation, retargeting optimization, and conversion funnel improvements.

Impact Identified TikTok and Shopee Ads as high-performing channels, uncovered inefficiencies in Google Ads and younger audience segments, and provided actionable strategies to improve ROAS and overall marketing profitability.

Problem Many Gen Z in Indonesia face financial vulnerability due to low financial literacy and weak financial planning.

Solution This project builds a data-driven dashboard to measure and visualize Gen Z financial readiness using the Financial Maturity Index.

Impact The dashboard analyzes financial readiness using 5 key indicators and 100+ survey responses, enabling clearer insights for financial education strategies.

Problem FMCG companies face demand uncertainty that often leads to stockouts or overstocking due to the lack of reliable data-driven demand forecasting.

Solution Built an ARIMA-based forecasting pipeline with data preprocessing, outlier handling, Box-Cox transformation, and stationarity testing to predict daily sales demand.

Impact The model provides short-term demand forecasts that support inventory planning, improve supply chain efficiency, and enable more data-driven operational decisions.

Problem South Sumatra faces low school participation and rising dropout rates among school-age students.

Solution This project builds a data-driven dashboard to analyze dropout patterns and education participation indicators.

Impact The dashboard analyzes 3,000+ dropout cases and APS indicators, helping identify high-risk areas for targeted education interventions.

Problem Businesses struggle to identify and differentiate high-value and inactive customers, resulting in inefficient one-size-fits-all marketing strategies.

Solution Applied RFM Analysis and K-Means Clustering to segment customers based on purchasing behavior.

Impact Successfully identified 2 customer segments, including high-value customers and dormant customers, enabling targeted retention and reactivation strategies.

Problem Manual interpretation of cardiac ultrasound images is time-consuming and prone to subjectivity.

Solution A CNN-based VGG16 model is implemented to automatically classify VSD conditions from ultrasound images.

Impact The VGG16 model achieved strong performance with 89% accuracy and 91% precision in classifying heart disease images.

Problem Revenue drops and product issues are often detected too late due to manual monitoring and lack of structured prioritization.

Solution A rule-based decision intelligence system that converts retail data into KPIs and automated role-based alerts.

Impact Enables earlier risk detection, prioritizes critical issues, and supports faster data-driven decision-making.