Assist with implementation of Reinforced Machine Learning/Kalman/something to derive optimal weights to X variables that are weighted based on forecastability of Y

I want to see how I can use Reinforced Machine Learning, Kalman Filter or something that penalizes and rewards X variables based on their ability to forecast a Y variable. I have already specified the X and Y variables: X variables : Trade Balance, PMI, Unemployment Rates etc. Y variable : Coincident Indicator I would like to determine how much to weight (sum to 100%) to the X variables such that variables are weighted more or less based on their forecastability of Y. Thought this could be something that reinforced ML or Kalman filter could assist with open for suggestions. I am looking for matlab code that implements the above. I have excel file that has all the data & mcode and explains a basic workflow using assumed optimal weights.