Data points from data sources are needed for today's forecasts to predict future events accurately. There is a need for the data to be complete and accessible without missing points that interfere with more inferior results. A solution for filling in missing points is proposed in an automated way using the k-mean algorithm, which allows obtaining the necessary data from the nearest data sources. To estimate that these data are geographically close to each other, k-means use the clustering principle by dividing the data sources into specific clusters, respectively, taking their location by Lat and Lng. The existing solution helps to use AODPF with previously missing points, which was not possible when repeating the experiment with those metrological stations for which it was not possible. Article answers to three questions: Does the missing data fill in increase the accuracy of the forecast? The algorithm that fills the missing data shows automation and standard features? Does the Kalman filter increase the accuracy of the forecast? Of course, with limitations.