We are developing global Artificial Intelligence to predict stock and currency markets.
Feel free to participate if you are the person with strong machine learning and AI skills.
Questions and Answers
What if I have no experience in predicting finance?
Feel free to apply anyway. Because you will not need to collect or structure the data. All the tasks will be in easy-to-work format.
What the winner will get?
The first 10 places will be awarded according to minimum logloss of the model. The amount of awards is still undisclosed.
When tournament will start?
The tournament will begin Spetember 25, 2017. And we will accept new prediction uploads until November 25, 2017. Winers will be rewarded at December 5, 2017.
How results will be calculated?
As soon as you upload your predictions we will show your logloss calculated on 30% of predictions. As well you can find yourself in leaderboard. The final logloss will be calculated on anther 70% to prevent overfitting.
Should I disclose my model?
You can retain all intellectual property rights to your model but we can offer you to buy licence to use your model in addition to the reward. As well you can get job offer in case of your interest.
What data will be like?
Most likely there will be 3 or 4 tasks with different level of complexity. Each task contains: train.csv, test.csv, sample_submission.csv files. More details are coming.
Train your skills
If you would like to train your skills before start of the main tournament we have interesting task for you.
We analyse a lot of data from professional traders. What if we can predict probability of profitable trade at the time when trade was placed? In the dataset we have some information about trades:
unique trade id.
unique trader id.
unique symbol id.
action of the trade. 1 if buy, -1 is for sell
ordered number of day of the week. 1 is for Monday
ordered number of hour of the day
count of opened trades that given trader had at the time of placing trade.
percent of gain for the trade, field we need to predict
You can find out submission format in Decision Tree Regressor sample code that have 9.7036 Mean Squared Error. If you can beat the results, please send your predictions to firstname.lastname@example.org we will email you back when your MSE will be calculated.