An ML specialist must forecast daily sales for one store using 10 years of historical daily sales data. About 10% of days in the historical record have missing sales values. The forecasting model underperforms. Which action is most likely to improve model performance?

Filling missing daily sales values using linear interpolation is most likely to improve model performance because forecasting models, especially those for time series data, perform poorly with gaps in the input data. Linear interpolation estimates missing values based on the values before and after the gap, providing a continuous dataset that models can process effectively. Aggregating sales across stores is not relevant as the problem specifies forecasting for one store. Applying smoothing for seasonal variation might be useful but doesn't address the fundamental issue of missing data. Changing the forecast frequency to weekly would reduce the granularity of the forecast and might mask patterns, rather than directly solving the data incompleteness problem.