Shu-Chu Liu

Accurate crop yield prediction is fundamental to sustainable agricultural management, enabling optimized resource allocation and informed decision-making. However, a critical gap exists in current prediction models: existing approaches overlook the temporal alignment between meteorological conditions and production management phases—defined as the intervals between consecutive agronomic operations (e.g., sowing, fertilization, thinning). This oversight results in suboptimal predictive performance, as conventional whole-season weather aggregation fails to capture phase-sensitive crop–weather interactions. While machine learning (e.g., XGBoost) and deep learning approaches (e.g., CNN, LSTM) have been applied to yield prediction, these models typically treat weather variables as temporally homogeneous inputs, inadequately modeling the correlation between historical yields and phase-specific meteorological patterns. To address this gap, this study proposes CNN-LSTM-AM, an innovative hybrid deep learning model that integrates convolutional neural networks (CNNs), long short-term memory (LSTM), and attention mechanisms (AMs), utilizing weather data explicitly aligned with production management phases as input. The CNN component extracts cross-phase weather patterns, the LSTM captures sequential dependencies across growth stages, and the attention mechanism dynamically weights phase importance based on meteorological conditions. The proposed model is validated using a real-world case study of Bok choy production from an agricultural cooperative in Yunlin County, Taiwan, encompassing 1714 production cycles over eight years (2011–2019). Experimental results demonstrate that CNN-LSTM-AM achieves an RMSE of 1448.24 kg/ha, MAPE of 3.60%, and R² of 0.98, outperforming five baseline models—CNN (RMSE = 2919.18), LSTM (RMSE = 2529.74), CNN-LSTM (RMSE = 1516.44), LSTM-AM (RMSE = 2284.64), and XGBoost (RMSE = 3452.47)—representing a notable reduction in prediction error (58% lower RMSE) compared to XGBoost. Furthermore, prediction accuracy improves progressively as harvest time approaches, and phase-specific weather encoding enhances accuracy by 16.5% compared to whole-season averaging. These findings underscore the critical importance of integrating agronomic domain knowledge into data-driven prediction frameworks.

Making an accurate crop harvest time prediction is a challenge for agricultural management. Previous studies of crop harvest time prediction were mainly based on statistical methods, and the features (variables) affecting it were determined by experience, resulting in its inaccuracy. To overcome these drawbacks, the objective of this paper is to develop a novel crop harvest time prediction model integrating feature selection and artificial intelligence (long short-term memory) methods based on real production and climate-related data in order to accurately predict harvest time and reduce resource waste for better sustainability. The model integrates a hybrid search for feature selection to identify features (variables) that can effectively represent input features (variables) first. Then, a long short-term memory model taking the selected features (variables) as input is used for harvest time prediction. A practical case (a large fruit and vegetable cooperative) is used to validate the proposed method. The results show that the proposed method (root mean square error (RMSE) = 0.199, mean absolute percentage error (MAPE) = 4.84%) is better than long short-term memory (RMSE = 0.565; MAPE = 15.92%) and recurrent neural networks (RMSE = 1.327; MAPE = 28.89%). Moreover, the nearer the harvest time, the better the prediction accuracy. The RMSE values for the prediction times of one week to harvesting period, two weeks to harvesting period, three weeks to harvesting period, and four weeks to harvesting period are 0.165, 0.185, 0.205, and 0.222, respectively. Compared with other existing studies, the proposed crop harvest time prediction model, LSTMFS, proves to be an effective method.

Tilapia (Oreochromis spp.), a prominent warm water food fish, is one of the major fish species grown in the aquaculture industry in south-east Asia. Tilapia can tolerate adverse water quality and other stressors, like diverse salinity and fluctuation of pH value, better than most other commercial aquaculture species. Environmental fluctuations are one of the main factors that affect the outbreak of infectious diseases in cultured tilapia. Cultured tilapia in Taiwan appears to be more susceptible to infections caused by Streptococci during the summer season. The present study emphasizes the Streptococcus spp. infection in tilapia in Taiwan and is the first study on the analysis of the potential impact of climate change on streptococcal infection in cultured tilapia in Asia. The data collected from the treatment and diagnosis system (TDS) of the aquatic animal diseases database from 2006 to 2015 were used to analyze the endemic streptococcal infection and the effect of climatic factors. Based on the results, the factor, average atmospheric pressure, is negatively correlated to streptococcal infection, while the other three, including average temperature, ultraviolet (UV) index, and rainfall, are positively correlated to streptococcal infection. A multivariate logistic regression model with these four factors was also built. When the average temperature is above 27.0 °C, the average atmospheric pressure is lower than 1005.1 hPa, or the UV index is above 7.2, the percentage of cumulated positive farms from all submitted tilapia cases was more than 50%. In addition, within 3 days of rain, rainfall is relevant to the occurrence of Streptococcus in tilapia. Using TDS to alert the occurrence of streptococcal infection in tilapia can be a very useful tool for veterinary aquatic animal inspection stations, and reducing economic losses and labour costs in aquatic agriculture.

In the past, ordinary service (single service) for computer repairs with the same service time and price based on the maximal profit criterion cannot satisfy the demand for priority service. In addition, the factor of customer satisfaction is always ignored in the service model. In this paper, a new bi-objective differential service model taking customer preference information into consideration is proposed for customer relationship management. In addition, a heuristic method is proposed for the service model since it is a combinatorial problem. The results show that the exhaustive enumeration search for the bi-objective differential service model is better than other methods for the single and differential service models. Furthermore, compared with the exhaustive enumeration search, the proposed heuristic method performs effectively and efficiently. Some sensitivity analysis is performed as well.

Numerous heuristic methods have been proposed for the fleet size and mix vehicle routing problem (FSMVRP), since the FSMVRP is an NP-hard problem. However, the vehicle type information is always ignored in the local improvement approaches of these heuristic methods. Hence, the solutions found by these heuristic methods without the vehicle type information are worse than those with the vehicle type information. In this paper, a hybrid heuristic method, incorporating the vehicle type information into variable neighborhood search (VNS), is proposed. The experimental results indicate that the proposed method is better than other reported methods and VNS without the vehicle type information in terms of average percentage deviation.