Fruit Market Trend Forecast Using Kmeans-based Deep Learning Models

Abstract

The fluctuations of fruit market price are mainly related to the fruit output quantity that may be influenced by climate, pest, and many other natural disasters. In this paper, in order to precisely forecast the coming trend of fruit market, image clustering-based deep learning framework is proposed. Initially, a series of data points indicating fruit prices are transformed into a series of fixed-length two-dimensional curve images at intervals, and each image is segmented into the input curve and the output curve. Furthermore, to make training set, K class labels are obtained on the output curves using the Kmeans clustering. Finally, the training set are employed for training convolutional neural network, long short-term memory and the hybrid of convolutional neural network and long short-term memory. The comparative study shows that the convolutional neural network has more advanced capability in predicting the fruit market than the other two, while the prediction accuracy of these trained models may not be sufficiently high.