The yield forecasting can also achieved by image analysis and machine learning. We can calculate the leaf area of the cucumber plant and further estimate the leaf area index by using the self-developed facility crop leaf area real-time monitoring and environmental sensing system, along with Python language to write leaf area analysis and calculation software, as well as Ndimage, Polyfit, Scipy, Polyval, and other packages to perform image output, linear fitting, calculation, and regression equation output. Then, the yield prediction is carried out in a regression model using random forest machine learning. To achieve more accurate predictions, the inputs of established model includes environmental parameters related to yield and leaf area. The accuracy of model was evaluated with different indicators (RMSE, MAE). The established model has been appropriately applied to the yield prediction of cucumber, following a week (7 days later). With the increased amount of information collected, we look forward to improving the accuracy of yield prediction and number of predictable days in the future to provide better reference data for growers. This will enable growers to further master product supply capacity and status, formulate appropriate management models and effective strategies, and visually display the forecast data of the next week's yield through the website loading the model of Flask platform architecture so that growers can quickly grasp trend changes and then make relevant decisions.