Topcoder Cognitive CommunityThis challenge is part of Topcoder Cognitive Community challenge series. Along with cash prizes, you will win Cognitive points towards the Leaderboard and the leader at the end of the challenge series wins an all expenses paid trip to TCO17.
The placement you earn in these challenges (non-F2F) will determine how many Cognitive points will be added to your total on the Leaderboard:
1st place = 500 pts
2nd place = 350 pts
3rd place+ = 100pts
If you have not already, go to http://cognitive.topcoder.com/ and join the Topcoder Cognitive Community.
Challenge ScopeIn this challenge you will prepare services for
- IBM Bluemix Weather Company Data
- IBM Bluemix Retrieve and Rank
And you will use them to implement the following staff:
1. Pull historic weather data / forecast for the specified locations / periods of time.
2. Based on historic weather data generate mock sales data and upload them to Retrieve and Rank.
3. Use Retrieve and Rank to pull historic data / forecasts / alerts.
Weather DataWe want to use real weather data all around. This seems to be quite straightforward, we just need to prepare a convenient service wrapper around IBM Bluemix Weather Company Data for getting weather history / forecasts / alerts. In case of any doubts don't hesitate to ask about details in the forum.
Sales DataWe will mock sales data. As we want to further use them to demostrate some machine intellegence in hihgligthing correlations between sales and weather, we should generate mock data in somewhat smart way. While the actual implementation is up to you to elaborate, here is the rought idea:
1. We configure a set of locations we are working with, and a timespan for generated data.
2. We configure a list of sale articles, for each article config will include:
- average sale rate (like items sold per average day)
- whether the good or bad weather benefits / spoils / have no impact on sales of this item.
3. You pull real wether data for locations / dates from (1), and for each day, location, item you generate the sold amount based on weather and config from (2). You use random function to make generated data somewhat fluctuating, but they should follow the config, i.e. the average sale rate should be achieved, on average, for days with moderate weather, somewhat worse / better sales should be achieved on days with better / worse weather.
Again, feel free to ask about details in the forum.