I am looking for someone with Light GBM / Catboost knowledge and who knows the ins and out of setting custom evaluation metrics, custom losses etc. If you do not know in detail about this please do not waste your time or mine with bidding/applying.
Making a traditional model you would have RSq, MSE, AUC etc as custom evaluation metrics, and binary accuracy for classification problems.
What I am looking for is a custom metric, which we can call “profit”. The dataset enclosed to this project the example dataset to be used. It is a list of different investment cases. If they are positive (1 in Win column – ie that case is the “winner”) the profit is in column “Return”. If it not true the loss would be -1 for that row. There can be only one investment case per groupid column (ie only one of the ones with the same groupid can be “the winner”), so if the model selects the wrong one a loss of -1 will occur. The column “Return” should only be used to calculate the profit, and not as a feature input in the model.
I am looking to maximise the total profit during eval / optimization. The total would be calculated on as the sum of profit of the models selected cases (The “winners”), and this is the number we would like to maximise.
To summarize I would like to have custom loss, metric and eval based on maximising of profit. The model should be Light GBM or Catboost, and use groupid to rank the investment cases by group.
For someone that knows their ML this project should only take a short time, but I am setting this as a per hour project to possibly extend the project with the developer for other features after the first task is finished.
9 freelancers are bidding on average £14/hour for this job
Hi, how are you? I'm data scientist specializing in: R / Python / Spark / Hadoop/ TensorFlow/ SQL / PowerBi. I would like to know more in detail about the project by private. Thanks.
Hi, I have +5 experience dealing with machine learning algorithms and worked on multiple projects in this field, I absolutely can do your project as you like. Please contact me to discuss more. Have a nice day