My Solution - score 1080


Hi everybody,

my two cents worth on the Big Mart Sales 2 competition can now be found on GitHub:

Not particularly stellar, but good enough for place 37 in the final ranking. Any comments and (constructive) criticisms are welcome. Let me know what you think!


How to improve score

Very nicely done :slight_smile:


Very well explained…Thnx alot


Excellent work!


Really Nice!!


Thank you. Great work !


Thanks Michael!
Your solution was very helpful. Now, I am more confident to take part in competitions.


Solution is awesome! I like it.

However, i am wondering if we can use this model to draw any inferences what so ever? Neural nets or Deep learning algorithms are suited for NLP, Image classification etc where in we don’t really want to find out the underlying predictors. However, in this situation, problem states that
"Using this model, BigMart will try to understand the properties of products and stores which play a key role in increasing sales."

So can we use your model to answer the problem?



Hey you computed the visibility values which sums up to 100 using test and training dataset but there is private test dataset as well so how then will you make that sum to 100 in private dataset


Hi Michael,

Nicely done. I’m relatively new to R and ML and your solution will allow me study and learn to be a better student of ML.


Firstly. Very impressive work.
Secondly,To count “count fat levels for each Item type” you could have used CrossTable function instead of the complex query (useful for people with experience in SAS)