Extracting association rules in relation to precipitation and effective factors

In recent years, knowledge production from the massive amount of data using data mining techniques has attracted attention. Meanwhile, prediction of precipitation in various hydrological issues such as runoff, flood, and drought as well as watershed management is of great importance. Accordingly, the purpose of this research is to extract association rules using data mining techniques to verify and predict the amount of precipitation. The monthly data of precipitation and effective factors related to it were used in this study. This research was carried out in Qazvin Plain for 30 years from 1988 to 2018. Different factors affecting the amount of precipitation with different intervals, including time without delay and delay of 1 to 3 months, were used. Four scenarios were defined based on the four timescales of the influential factors. For each scenario, rules on precipitation and its influential factors were extracted by the Apriori algorithm. The extracted rules were evaluated by the indicators of confidence, support, and lift. The accuracy of the rules was evaluated for all four scenarios according to the three indicators and the best scenario was chosen. According to the results of the evaluation indicators, it was determined that effective factors with the 2-month delay had the most substantial effect on predicting the amount of precipitation. In the last step, the independent relationship between precipitation and factors affecting the 2-month delay was examined. Finally, it was determined that the average pressure level factor of the station with a 2-month delay had the most significant relationship with precipitation in Qazvin Plain.