"There's A Better Way To Do": My Appalling Welcome To Kaggle

After three weeks of learning at Data Science Nigeria, we felt we were ready for the world, the competitive one we're in. Kaggle was the place to go, where you've got knowledge based competitions for beginners, I'm still in this group I guess, well paid competitions for those who've known their ways about the game, and finally in-class contests, which I participated in.

Registering, I'd say, is the easiest thing you can do on Kaggle, as every moment after that could be defining - whether you're a winner or not. I registered along with my colleagues, anticipating what was ahead, my first contest as a programmer.

The competition finally began and the data sets, which sometimes could be confusing, were available for download. I downloaded mine, submitted the sample submission to know how much work was there to do, and to my surprise, an accuracy above average was there. This made me relent as I felt all I needed to do was just a little modification, and to make me more complacent, I had an accuracy of about 80 per cent, putting me as high as seventh on the leaderboard when arranged alphabetically and third when arranged based on accuracy.

I forgot that "There's a better way to do everything", some people didn't forget. They improved their submissions repeatedly, climbing up the leaderboard gradually, bringing me and some other forgetful ones down. It was shocking! I was now down, twelfth of seventeen, I made the first 71 per cent you'd say. It's not such a good start to me.

The next contest was different, I was determined and fully prepared - or so I thought, to win this. I did the usual, downloaded the data sets, submitted the sample submission - I got a very low score this time, and began from scratch. The last defeat - that's what I'll call it, served as my main source of determination along with other factors. I read my data sets, observed them, made them understandable to python - I just turned non-numeric data to numbers, and my base model was ready. I submitted it and what a meteoric rise to the top I had, my new accuracy was like six times the previous one, this was going to be fun I thought.

I decide to use a better way, improve my model, the improvements made were tardy, I started losing patience, I was not used to this, but I remained adamant to win. I started my feature engineering and after a couple of attempts, I had a positive result, though not large, it was an improvement. I submitted and I was in the fifth position, with a respectable margin from the sixth.

I logged in the next day to discover that I've dropped to sixth position, I was angered by this, I made new models and I moved to fifth position again, I guess I was happy again. I relented.

The contest finally ended with some deadline submissions, that sent me to the bottom again, I was eleventh of twenty-four, I guess I still have a long way to go.

I hope I can go the distance and be a winner soon.