Impact of Missing Data Imputation Methods on Univariate Turkey Production Time Series Analysis and ARIMA-Based Forecasting

missing VALUES, which are the absence of observations for certain variables within a dataset, occur in time series data and can disrupt continuity and lead to inaccurate predictions. Ensuring the accuracy and reliability of time series forecasting results necessitates the proper management of these missing values. Handling missing values in time series data poses a considerable challenge when utilizing the ARIMA model. This research utilized annual turkey production data spanning from 1968 to 2021. The dataset exhibited outliers and missing values. These gaps indicate a systematic pattern of missingness, likely classified as Missing at Random (MAR), which necessitates effective handling strategies for trustworthy forecasting. Numerous strategies have been suggested to tackle these challenges, including data deletion and imputation techniques. Two techniques were used in this study to estimate the dataset's missing values and their performance was compared. The two techniques used were the linear interpolation method and the mean series method. To assess the goodness of fit for both approaches, performance metrics including root mean squared error (RMSE), mean absolute error (MAE), mean square error (MSE), mean absolute percentage error (MAPE) and coefficient of Determination (𝑅 2 ) have been calculated. The findings of this study demonstrate that the linear interpolation method is significantly more effective at imputing missing values compared to the mean series method, as evidenced by R² value of 0.904 versus 0.213, respectively. Missing values in time series data disrupt continuity and lead to inaccurate forecasts if not properly addressed. The results indicate that linear interpolation better preserves the temporal structure of the data, resulting in improved imputation quality and more reliable predictive performance than mean series imputation.