Emerging IT Skill Demand Analysis and Skill Trend Prediction Using Job Posting Data and Machine Learning

Abstract: Online job advertisements describe the skills employers request, but this information is embedded in unstructured text. This paper presents a data analytics framework that cleans job-posting records, extracts IT skills with a documented dictionary, measures demand as the share of unique postings mentioning each skill, identifies emerging skills from changes in that share, and tests whether machine learning can forecast next-month demand better than simple baselines. The framework is implemented and validated on a simulated corpus of 11,820 postings spanning 36 months (January 2023 to December 2025) with 30 skills. Rule-based extraction with context checks removed all 269 false matches produced by naive keyword matching on this corpus. In rolling one-step-ahead forecasting over a nine-month test period, a damped-trend exponential smoothing model gave the lowest error (MAE 1.49 percentage points), 18.0% lower than the naive baseline, while a Random Forest on lag features did not outperform a moving average. Because the corpus is simulated, the numerical rankings and trends demonstrate the method only and are not measurements of the real labour market; the same code is intended to be re-run on a documented public job-posting dataset.

Keywords: IT skills, job postings, text mining, skill extraction, machine learning, demand forecasting.