Analysis and Prediction of Wordle Dataset Based on ARIMA and Pearson's Correlation Coefficient

Abstract

In this paper, a time series analysis was performed to ensure the parameters p, d, and q. Finally, the ARIMA (1,1,0) model was chosen to create a prediction interval for the number of outcomes to be reported in the future. The results of the precision test showed good confidence in the prediction with a goodness-of-fit of 98.2%. the prediction interval for March 1, 2023 was [10103,10523]. In addition, this paper used Pearson's correlation coefficient and chi-square test to verify the correlation between each attribute of the word and the percentage of scores in the hard model. In summary, this paper found that the attributes of words definitely affect the percentage of scores, and these scores are mainly the initials and lexicality of words.