Prediction of Scrabble game players based on ARIMA and Machine Learning under big data

Abstract

Wordle is a favored puzzle. Firstly, to predict its daily player count, based on analyzing the inherent trend of the data and verifying its stationarity, the effectiveness of using the ????????????????????((2,3,5),2,(2,3,5)) model to predict the number of reported results is proved. Before predicting, this thesis preprocesses the data from two parts: Number of reported results and Number in hard mode. Then, the 95% prediction interval for March 1, 2023, is [8626 16199]. Secondly, to predict the distribution of reported results percentages, this thesis uses 7 random forest regressors, their feature variables have 5 dimensions (Contest number, Frequency of word, Number of repeated types, Number of letter types, Maximum number of repetitions), and the response variable takes one of the percentages (7 dimensions) in turn. Results show that the distribution of reported results percentages of “SALSA” on March 1,2023 is [0, 2, 15, 35, 31, 15, 2].