Correlation (CORR)
Last updated
Last updated
Correlation, in the stock or investment industries, is a statistic that measures the degree to which two securities move in relation to each other. Correlations are used in advanced portfolio management, computed as the correlation coefficient, which has a value that must fall between -1.0 and +1.0. This metric serves as a measure of the model's performance in forecasting by evaluating the correlation between our forecast series and the actual series. A higher correlation indicates improved portfolio management and greater residual returns.
It is essentially the correlation between the predictions made by the model and the corresponding label or return of price. The target value is determined during the Labeling phase and serves as the basis for calculating this correlation.
You are required to submit the correct result within our designated timeframe, specifically once every hour. However, if you are unable to meet the submission deadline, there is no need to worry. Our label already incorporates a weighted moving average (WMA) combination of the information from the past n rounds. This means that when you submit your entry, you will have up to n rounds before you need to provide your results again.
We will utilize your predicted series to calculate its correlation with the actual label. In case you miss a submission for any particular hour, we will use the results from the previous hour to fill in your forecasted results for the missed time period.
To provide additional clarity, presented below is a diagram illustrating the concept:
Yn: your submission file
X: missing submission for a specific hour, this will be replaced with the last submitted file for that hour.
CORR15 means we use the last 15 rounds to calculate
