The Farmer Exchange Rate (FER) is one of the indicators in determining the welfare level of farmers. Every month this large FER has a change especially at the end of each season during harvest, the price game especially from producers often occurs so that not a few farmers complain and suffer losses. Therefore, there need to be efforts from the government is looking at past data trends to find the amount of FER in the future so that policies can side with farmers. Artificial Intelligent with a multilayer Back Propagation method is very good for modeling and forecasting of data series in the past with data input in the form of a matrix mxn. The simulation results show that in 2020 in NTB there is an average increase in the exchange rate of farmers of 118.76%, this means that farmers are experiencing a surplus or farmers’ incomes are rising more than their expenditures. However, it appears that from June-August and October-November there was an average decrease of 2%, although overall there was increase of 2.076%. This result was obtained by the construction of a network architecture of two hidden layers where theMSE of 0.12, and MAPE of 0.23.

This study aims to predict the smoothness of installment  payments in cooperatives, making it easier for staff to analyze credit lending. Lack of prudence in analyzing credit results in customers who are in arrears in paying installments, resulting in bad credit. To minimize errors that exist, it is necessary to evaluate the provision of loans to prospective debtors. By utilizing past member criteria data in the past that will be used to predict smooth payments using data mining. The data mining technique used is the Naive Bayes classifier method. The prediction process uses the naive Bayes method, namely by determining the probability or opportunity based on the previous member's data, and the results are used to help make a decision. The criteria used are member data: employment, income, house status, number of credits, and type of credit. Based on the naive Bayes method, the results obtained are 90.00% accuracy, 0.880% AUC, 83,33% recall, and 100% precision.