PokeHack

• 

  Correlation matrix: After Cleaning

• 

  Correlation matrix: Before Cleaning

• 

  Data Visualization per user

Inspiration

PokeHack is built to provide predictions to the user about the wait time that he/she might have to experience in order to run any jobs which are currently being used by another resource. The inspiration was predicting wait time comes from the profound need for saving resource time.

What it does

If we can predict the wait time for each user for every job, then its possible for a system to recommend the best job that the user can execute and has the least wait time.

How we built it

We built the system in Python and performed data analysis with Tableau. Initially, we created the data set for each user by iterating through the given data set and performing filtering and choosing all the metrics of one pokemon. Once we had the dataset for each pokemon, we visualized the data on Tableau. After observing the data on Tableau, we observed the outliers like values where there were no jobs but still recorded wait time. So data pre-processing was done on python to normalize the data. Once a clean data set was found, the correlation matrix was plotted for all the metrics to find the dependency of each column and hence identify their relationship with the wait time. Beyond this wait time was defined as prediction column and t-test was performed. All insignificant parameters were removed and regression was performed to predict wait times. More machine learning algorithms like neural networks can be performed on the dataset to achieve higher accuracy in predicting the wait time.

Challenges we ran into

We ran into a lot of challenges in correlating the data but we were proud that we used various visualization and statistical techniques to discover dependency.

Accomplishments that we're proud of

We could demystify the hidden patterns in the data to make more sense and understand what lay within those multiple rows and columns with a clock ticking in our heads.

What we learned

We learned that interpreting the data could be tough in real life and the experience we gained in analyzing the data set and building something meaningful was useful.

What's next for PokeHack

We believe PokeHack gave us a good taste of what problem-solving in the real world feels like and we learned the areas that we could improve on. So next for PokeHack would be to find more efficient ways to achieve the solution.

Built With

• machine-learning
• python
• tableau