Time series analysis and forecasting of cholera disease using discrete wavelet transform and seasonal autoregressive integrated moving average model

Purpose: Cholera is among the leading causes of death in Nigeria. The main predictors of cholera transmission remain the lack of access to potable water and good sanitary conditions. Cholera is also linked to weather variables such as maximum temperatures, high Rainfall, and humidity. The relationship between cholera cases and weather variables depends on location, time, or season; hence, it is a time series dataset. This research aims to enhance the seasonal autoregressive integrated moving average (SARIMA) model by incorporating the discrete wavelet transform (DWT). Methods: This research proposed a novel approach to forecasting cholera using the SARIMA model by incorporating DWT as a dimensionality reduction technique and a K-means clustering algorithm for outlier detection. The enhanced model is termed the "Enhanced seasonal autoregressive integrated moving average" (ESARIMA). DWT is a good dimensionality reduction technique for time series data and extracts the best features for forecasting to have better prediction accuracy and minimal error. Result: The results show that ESARIMA (accuracy = 97%, RSS = 0.502) outperformed the existing model, SARIMA (accuracy = 91.61%, RSS = 0.60). Conclusion: Nigeria's weekly and monthly cholera outbreaks exhibit stochastic seasonal time series behavior that becomes stationary after the first seasonal differencing; hence, it could be forecasted with specific time series models. (c) 2023 The Author(s). Published by Elsevier B.V. on behalf of African Institute of Mathematical Sciences / Next Einstein Initiative. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)