Research on Prediction of Freshness of Frozen Yellow Croaker Based on PCL Neural Network

This study proposes a PSO-CNN-LSTM model to address the challenge of predicting freshness during the storage of frozen yellow croaker. Firstly, the data information of volatile gases related to the preservation process of large yellow croaker was recorded through sensors. Then, the collected dataset was strictly preprocessed to effectively eliminate the negative impact of abnormal data on prediction performance, ensuring that the data is suitable for subsequent neural network model inputs; To address the issue of limited data on volatile base nitrogen, polynomial fitting is used to make the data distribution smoother and improve the prediction accuracy of subsequent neural network models; Finally, a PCL neural network is created, and convolutional neural networks are introduced to improve feature extraction capabilities. Long short-term memory networks are then used for prediction, and the best hyperparameters of the model are obtained through PSO algorithm to achieve high-precision prediction of the freshness of frozen yellow croaker. After a series of experiments, it has been verified that the PCL model performs better than traditional deep learning models in multiple evaluation metrics, including convolutional neural networks and gated recurrent units. This not only proves the effectiveness of the PCL model in predicting the freshness of yellow croaker, but also highlights its enormous potential in practical applications.