karanjagota / ml-minihack--datahulk

My solution to 6 Hrs Mini Machine Learning Hack #Ranked 18 in the competition with ( 0.681 )

ML MiniHack (DataHulk)

My solution to 6 Hrs Mini Machine Learning Hack. I #Ranked 18 in the competition with ( 0.681 ) score.

Aim : The aim of this competition was to predict the price of given stock will move up or down on the next day.
Metric Used : Logloss
My Methodology : Simple 4 step solution

1. EDA
2. Impute missing values if any
3. Feature Engineering
4. Model Development and CV

Solution Info : Data Quality was excellent and most of the variables which are important for predicting stock prices like volume , moving averages of various past days were alredy given. I decided to make simple gbm model with one feature engineering and considering data of only last 365 days. I wanted to optimize more by developing a neural network but due to short timespan couldn't. I finally end up with 18th rank on private leaderboard.