DATA MINING TECHNIQUES FOR DECISION TREE-BASED WEATHER PREDICTION

Resumen

Forecasting weather conditions is an important field of application in meteorology and one of the most challenging experimental and technical problems in the field. Here, we consider the possibility of using information mining techniques to evaluate features such as the highest and lowest recorded temperatures. This was accomplished using Choice Tree computations and meteorological data collected from the different metropolitan locations during the years 2012 and 2015. Variations in the climate that defy simple prediction provide a challenge to the methods used to predict them. Extremes in temperature, humidity, and wind speed are just a few of the climatic parameters whose limits are notoriously hard to pin down and evaluate. The Decision Tree Algorithm is used to remove embarrassing material from publicly available datasets. Maximum and minimum temperatures are the most important determinants of the weather forecast. We expect full cold or full heat or snowfall on the level of these limits. This research develops a model based on a decision tree that can predict extreme weather events including complete cold, full blistering, and snow precipitation, information that might save lives.

The variability of past conditions may help us predict the weather in the future. It's quite unlikely that the weather on the day you're thinking about will be identical to the weather on the same day a year ago. However, the chances of a match during the next two weeks of the previous year are quite high. Thus, a sliding window of length about seven days is selected for the two weeks evaluated during the prior year. After that, we match the mental week of the current year with each seven-day sliding window segment. Atmospheric forecasting includes having the optimal window available for use. Using a sliding-window method of analysis, a prediction is formed. Every month's worth of data is recorded for a lengthy period of time so that accuracy may be scrutinised in depth. Based on the results of the approach, it was concluded that the technique used to anticipate weather conditions was very effective, with an average accuracy of 92.2%.