I got to spend some time with my son this past summer.  He’s growing up and has his first job this year so my opportunities for spending blocks of time with him are dwindling.  He is a budding cinephile so one of the way we connect is discussing movies.  

While he was down here we started discussing “the best movies” and what the criteria for inclusion might be.  I was recalling that conversation last week and decided to try and do some data collection and machine learning to create a Top 100 Movies of All Time List.

I collected reviews, ratings, data sets – all the data I could get my hands on that had numerical representations of “movie scores”.  I normalized that data and then ran it through some machine learning techniques that I have been studying the last few months.

The Top 100 Movies of All Time:

The ratings have been rounded for display purposes but the movies are in the right order.  I tried to find a mix of data ( 1-10 stars, 1-100 percent ranking, thumbs up or down); from  critics and movie goers.

The biggest part of this project was collating and cleaning data.  It is not as sexy as Machine Learning but just as important.  Without clean data and quality feature engineering even the best deep learning algorithms are not particularly useful.

If you have data you need cleaned up…