Research on Percentage Prediction of Wordle Results Based on Weighted CLR-XGBoost
Feifei Ma, Qingkang Ni · 2023
For the guessing game with word text as the object, the contestants limit the percentage of the number of attempts in the process of solving the problem, and some decision-making suggestions can be fed back to the game developer. This paper proposes a wordle result percentage prediction and evaluation model based on weighted CLR-XGBoost.firstly, weighted CLR was used to transform the component data to eliminate the effect of multicollinearity in the percentage of trials. Then, based on the weighted CLR transformation data, the XGBoost report result prediction distribution model is established to determine the relevant percentage distribution on March 1, 2023. Next, based on the feature extraction of the uncertainty factors of the model based on EERIE words, it can be considered that Linsear Write Formula and Flesch Reading Ease formula have an impact on the model. Finally, the paper explored the credibility of the model and found that the R2of the training set and the test set exceeded 0.97, and the fitting effect of the model was significant.