Big Data Analytics Framework for Process Optimization and Yield Improvement in Semiconductor Production

Abstract

Semiconductor production generates large volumes of data from equipment sensors, manufacturing systems, wafer inspection, maintenance records, and final testing. This study develops a big data analytics framework to improve process optimization and wafer yield. The framework combines multiple production data sources, cleans and standardizes the information, and applies correlation analysis, random forest, and gradient boosting. Process variables such as temperature, pressure, gas flow, equipment alarms, and maintenance delay were evaluated for their effect on production quality. The results showed that average wafer yield increased from 89.4% to 95.2% after implementation. Process variation decreased from 8.1% to 4.6%, defect rate fell from 7.3% to 3.8%, and equipment stoppage declined from 6.5% to 3.1%. These findings show that integrated analytics can detect problems earlier, support better engineering decisions, and improve production stability while reducing avoidable wafer losses and delays. The framework offers a practical method for data-driven yield management in semiconductor manufacturing.