crab_house_synthetic_2025_to_2026

The dataset used in this study was developed to investigate the feasibility of predicting mangrove crab health conditions based on water quality parameters. In crab aquaculture, obtaining large-scale labeled datasets is challenging because field measurements are time-consuming, expensive, and require continuous monitoring under controlled environmental conditions. Consequently, only a limited number of real-world observations were available for this study. To address this limitation, the dataset consists of a combination of seven original laboratory observations and fifty-seven synthetic samples, resulting in a total of 64 observations. The original data were collected from laboratory measurements conducted as part of an IoT-based water quality monitoring project for mangrove crab cultivation. The synthetic samples were generated using biologically plausible ranges of water quality parameters reported in the aquaculture literature, including temperature, pH, dissolved oxygen (DO), salinity, ammonia, nitrite, and turbidity. Label assignment (Normal, Stress, and Molting) was performed by considering the combined influence of these parameters based on established biological knowledge rather than random sampling. The synthetic data were introduced solely to augment the limited real observations and support preliminary model development. Therefore, the dataset should be regarded as a proof-of-concept dataset for evaluating the feasibility of machine learning and Explainable Artificial Intelligence (XAI) in crab health prediction, rather than a representative dataset for large-scale deployment. Future work will focus on collecting substantially larger real-world datasets to validate and improve the proposed framework.