Predicción del índice de madurez del pepino (cucumis sativus l) mediante la aplicación de la inteligencia artificial

Abstract

Cucumber (Cucumis sativus L.) is a globally consumed vegetable and is highly valued in the food industry. However, its quality largely depends on its physicochemical properties, such as texture, soluble solids, pH, and acidity, which are gradually affected by internal and external factors, including temperature, humidity, and storage time. This study aims to predict the physicochemical changes in cucumber (Cucumis sativus L.) through the application of machine learning. The methodology involved physicochemical analyses of the vegetable and image capture for maturity determination. The study was conducted using Google Colab with Python, leveraging specialized libraries such as TensorFlow and Keras, along with complementary tools from scikit-learn and Matplotlib. Deep neural networks with convolutional layers were implemented for feature extraction, incorporating pooling techniques to reduce image dimensions. The model was trained over 30 epochs, with hyperparameter tuning to optimize accuracy. Model performance evaluation was conducted using a confusion matrix, enabling the determination of the effectiveness of the algorithm in classifying the maturity index (IM, M). To statistically validate the results of the chemical properties, a T-student test was applied for independent samples with a significance level of α = 0.05, establishing a cutoff point based on a pH value of 4.68. The findings demonstrate the feasibility of using machine learning to monitor cucumber quality, providing a data-driven approach for optimizing post-harvest management.