Need Advice on Improving Keyword Ranking for My Tool

Hi everyone, I’m building a tool that scrapes articles for a search keyword, extracts related keywords, and scores them based on occurrences (title, headers, body) and semantic similarity. I then compare my extracted keywords to those from Surfer SEO and Clearscope, achieving \~60% overlap. My challenge: I can’t effectively re-rank my keywords to ensure the 60% overlap appears in my top 100 results. Do you think my approach is solid? How would you improve it? I’m considering training a tabular ML model for re-ranking—what parameters would you recommend? I’m new to this, so any guidance is appreciated!