Farmer suicides have reached a concerning level in India recently. This issue mainly stems from farmers’ inability to sell their products at the desired profit level, which is caused by price fluctuation in the agriculture market. To help the farmers with this issue, this paper proposes a deep learning algorithm, PECAD-CLS, which can predict the future crop price trends (Increase, Decrease, Stable) based on the historical patterns of crop price and volume. Even though previous studies have attempted to tackle market price trend prediction via Machine Learning (ML) algorithms, they do not model the spatio-temporal dependence of future prices on past data explicitly. Hence, they do not have a desirable performance on the spatio-temporal datasets. To address this deficiency, our proposed method makes two main contributions. First, we collect real-world daily price and volume of different crops over a period of 11 years and then impute it to deal with missing values. Second, we modify a state-of-the-art model, called PECAD, to predict the future produce price trends. Our experiment results show that PECAD-CLS improves state-of-the-art baseline models by ∼5% in terms of F1 (in the best case scenario). In addition, the PECAD architecture performs even better in direct crop price prediction; it outperforms baselines by ∼24% in terms of coefficient of variation.