Classification of Banana Packaging Type Based on Physicochemical Properties Using Multilayer Perceptron (MLP) Artificial Neural Networks

Pourdarbani, Razieh; Jafarzadeh, Sajad; Moradi, Farid

doi:10.22034/jrmam.2026.14799.736

Classification of Banana Packaging Type Based on Physicochemical Properties Using Multilayer Perceptron (MLP) Artificial Neural Networks

Document Type : Original Article

Authors

Razieh Pourdarbani

Sajad Jafarzadeh

Farid Moradi

Department of Biosystem Engineering, University of Mohaghegh Ardabili, Ardabil, Iran

10.22034/jrmam.2026.14799.736

Abstract

Abstract
The neural network technique is a powerful modeling method that mimics biological memory and learning using neurons. In this study, the classification of packaging types under dynamic and quasi-static loading conditions on the banana fruit properties were investigated. The input data for the neural network included the phenolic content, antioxidant capacity, acidity, pH, soluble solids, firmness, and moisture content. The classification output data consisted of the different packaging types were used in the experiment. A perceptron network with 3 and 5 neurons in different layers was employed, utilizing the hyperbolic tangent as the selected activation function. In this research, the performance of multilayer artificial neural networks for classifying and predicting experimental data were evaluated. The models were trained with varying numbers of learning epochs and iterations, and the results were assessed based on mean squared error (MSE) values. The best performance was achieved with the MSE of approximately 0.00019 for the training data and 0.02 for the validation data, indicating the high accuracy of the model under optimal conditions. Additionally, increasing the number of learning epochs and iterations generally led to reduced error and improved prediction accuracy. However, in some cases, extremely low training error values suggested the possibility of model overfitting.

Introduction
More than one - third of agricultural crops are wasted after harvest that leads to lack of food. Improving storage methods, especially in developing countries, can reduce waste, increase food security and support farmers. The mechanical properties of the material are more comprehensive and clear understanding of the physical properties and consequently the texture. Employing this knowledge is useful to reduce the damage during the required operation and to increase the shelf life of fruits . Hence, attention to the shape and physical dimensions are required to be used for separation and sorting. some of physical properties such as shape, size, surface area, density, porosity, color, and appearance are important in many problems depending on design of special machines or analyzing product behavior in material handling. Also, the appearance characteristics of fruits are affected by their value in the market. Hence, it is important to properly control the harvest and post-harvest operations. In recent years, several studies have been conducted to perform a precise evaluation of food products due to increasing consumer demand and their special attention to the internal quality of fruit such as freshness, sweet and nutritional value. There are various researchers conducted by differenet researchers. Azadbakht et al, (2019) conducted a research on the classification of chemical properties of pears under dynamic loading using artificial neural network .the results showed that radial basis function network had the best performance in the classification of loads .fu et al. (2016) conducted a study to classify kiwifruit on the basis of shape using a image processing. They measured length , maximum diameter of basal area and projected area of fruit. The stepwise linear regression was used to select important variables in predicting the minimum diameter of basal area and volume. Hosnavi et al (2011) investigated physical-chemical properties to classify date palm fruits in 14 native cultivars from different areas of morocco , tunisia and algeria .the results showed that sugar content are the dominant part of date fruitss .Also , the amount of protein , fat and ash were observed in the samples .
The aim of present paper is classification of banana fruit by using its constituent components in static and dynamic loads using MLP neural network.
Material and Methods
Based on the opinion of experts from local markets in Gorgan, a sufficient number of semi-ripe bananas were procured. First, the defective samples were separated. Then, the bananas were sorted by size to make them uniform in size and weight, thus eliminating the fruits from the standard size. Then, the fruits were packaged and stored for 14 days. The chemical properties of the samples were measured daily.
Results and Discussion
The models were performed in 2 period of learning and 2 replications for performance analysis. the overall results showed that increasing the number of learning period caused error reduction and performance improvement, especially in class 2 which recorded the highest accuracy
In contrast, Class 3 and Class 4 still had high errors, indicating the need for further optimization.
Also , the R value was negative, suggesting a mismatch between model predictions and actual data .the increase in the number of model runs had a significant effect on accuracy improvement in some classes .For further optimization, it is suggested to use methods such as weights, selecting the optimal architecture of neural network and optimization algorithms such as Adam or RMS prop.
Conclusions
The results of the implementation of neural networks showed that the accuracy of the model is strongly influenced by the number of learning periods and the number of replications. The increase in the number of replications, reduced the mean square error (MSE). In the case of the number of learning cycles of 1000, the final error in cross-validation is higher, indicating the possibility of overfitting in low-complexity models.
On the other hand, implementing learning period of 2000 resulted in a significant decrease in educational error and also improved model performance in experimental data .in some cases , very low MSE values indicate that the model accurately learns the training data , but there is a need to validate the performance of the model over the new data to ensure that the model is appropriate .Adjusting the weights , reducing learning rate, and increasing the number of training data can help improve model performance and prevent the overfitting.
Author Contributions
Razyieh Pourdarbani: Conceptualization
Sajad Jafarzadeh: Writing
Farid Moradi: Methodology
Data Availability Statement
"Not applicable".
Ethical Considerations
It is confirmed adherence to ethical standards, including avoidance of data fabrication, falsification, plagiarism, and misconduct.
Conflict of Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper
Funding Statement
The author(s) received no specific funding for this research

Highlights

Akandi, S. R., Qhodsi, S. S., Minaei, S., Najafi, G., & Hashjin, T. T. (2018). Mechanical Properties of (Aloe v era L.) Leaf for Designing Gel Extraction Machines. J. Agr. Sci. Tech, 19(4), 809–820.

Azadbakht, M., Mahmoodi, M. J., & Vahedi Torshizi, M. (2019). Effects of Different Loading Forces and Storage Periods on the Percentage of Bruising and Its Relation with the Qualitative Properties of Pear Fruit. International Journal of Horticultural Science and Technology, 6(2), 177–188.

Azadbakht, M., Torshizi, M. V., & Ziaratban, A. (2016). Application of Artificial Neural Network ( ANN ) in predicting mechanical properties of canola stem under shear loading. Agricultural Engineering International: CIGR Journal, 18(5), 413–424.

Azadbakht, M., Vahedi Torshizi, M., & Asghari, A. (2019). Biological properties classification of pear fruit in dynamic and static loading using artificial neural network. Innovative Food Technologies, 6(4), 507–520.

Azadbakht, M., Vahedi Torshizi, M., & Mahmoodi, M. J. (2018). Determination of pear bruises due to a thin edge compression load by CT scan method. Innovative Food Technologies (JIFT). https://doi.org/10.22104/jift.2018.2842.1684

Azadbakht, M., Vehedi Torshizi, M., Aghili, H., & Ziaratban, A. (2018). Application of artificial neural network (ann) in drying kinetics analysis for potato cubes. CARPATHIAN JOURNAL OF FOOD SCIENCE AND TECHNOLOGY, 10(2), 96–106. https://www.cabdirect.org/cabdirect/abstract/19981100164

B. Khoshnevisan, Sh. Rafiee, M. Omid, M. Y. (2013). Prediction of environmental indices of Iran wheat production using artificial neural networks. International Journal of Energy and Environment, 4(2), 339–348.

Balogun, W. A., Salami, M. E., Aibinu, A. M., Mustafah, Y. M., & S, S. I. B. (2014). Mini Review: Artificial Neural Network Application on Fruit and Vegetables Quality Assessment. International Journal of Scientific & Engineering Research, 5(6), 702–708.

Bondet, V., Brand-Williams, W., & Berset, C. (1997). Kinetics and mechanisms of antioxidant activity using the DPPH. free radical method. LWT-Food Science and Technology, 30(6), 609–615.

Diels, E., van Dael, M., Keresztes, J., Vanmaercke, S., Verboven, P., Nicolai, B., Saeys, W., Ramon, H., & Smeets, B. (2017). Assessment of bruise volumes in apples using X-ray computed tomography. Postharvest Biology and Technology, 128, 24–32. https://doi.org/10.1016/j.postharvbio.2017.01.013

Fadiji, T., Rashvand, M., Daramola, M. O., & Iwarere, S. A. (2023). A Review on Antimicrobial Packaging for Extending the Shelf Life of Food. Processes, 11(2), 590. https://doi.org/10.3390/pr11020590

Fathi, M., Mohebbi, M., & Razavi, S. M. A. (2011). Application of Image Analysis and Artificial Neural Network to Predict Mass Transfer Kinetics and Color Changes of Osmotically Dehydrated Kiwifruit. Food and Bioprocess Technology, 4(8), 1357–1366. https://doi.org/10.1007/s11947-009-0222-y

Fu, L., Sun, S., Li, R., & Wang, S. (2016). Classification of Kiwifruit Grades Based on Fruit Shape Using a Single Camera. Sensors, 16(7), 1012. https://doi.org/10.3390/s16071012

Ganiron, T. U. (2014). Size properties of mangoes using image analysis. International Journal of Bio-Science and Bio-Technology, 6(2), 31–42. https://doi.org/10.14257/ijbsbt.2014.6.2.03

Kolniak-Ostek, J. (2016). Identification and quantification of polyphenolic compounds in ten pear cultivars by UPLC-PDA-Q/TOF-MS. Journal of Food Composition and Analysis, 49, 65–77. https://doi.org/10.1016/j.jfca.2016.04.004

Li, W. L., Li, X. H., Fan, X., Tang, Y., & Yun, J. (2012). Response of antioxidant activity and sensory quality in fresh-cut pear as affected by high O2 active packaging in comparison with low O2 packaging. Food Science and Technology International, 18(3), 197–205.

Liu, Y., & Ying, Y. (2007). Noninvasive Method for Internal Quality Evaluation of Pear Fruit Using Fiber-Optic FT-NIR Spectrometry. International Journal of Food Properties, 10(4), 877–886. https://doi.org/10.1080/10942910601172042

Martín-Diana, A. B., Rico, D., Barat, J. M., & Barry-Ryan, C. (2009). Orange juices enriched with chitosan: Optimisation for extending the shelf-life. Innovative Food Science & Emerging Technologies, 10(4), 590–600.

Massah, J., Hajiheydari, F., & Derafshi, M. H. (2017). Application of Electrical Resistance in Nondestructive Postharvest Quality Evaluation of Apple Fruit. Journal of Agricultural Science and Technology, 19, 1031–1039.

Mohammad Vahedi Torshizi, A. A., Tabarsa, F., Danesh, P., Ali, & Akbarzadeh. (2020). CLASSIFICATION BY ARTIFICIAL NEURAL NETWORK FOR MUSHROOM COLOR CHANGING UNDER EFFECT UV-A IRRADIATION. Carpathian Journal of Food Science and Technology, 152–162. https://doi.org/10.34302/crpjfst/2020.12.2.16

Mohsenin, N. (1968). Physical properties of plant and animal materials. Journal of Agricultural Engineering Research, 13(4), 379. https://doi.org/10.1016/0021-8634(68)90151-0

N. Galili, I. Shmulevich, & N. Benichou. (1998). Acoustic Testing of Avocado for Fruit Ripness Evaluation. Transactions of the ASAE, 41(2), 399–407. https://doi.org/10.13031/2013.17164

Pérez-Jiménez, J., & Saura-Calixto, F. (2015). Macromolecular antioxidants or non-extractable polyphenols in fruit and vegetables: Intake in four European countries. Food Research International, 74, 315–323. https://doi.org/10.1016/j.foodres.2015.05.007

Salehi, F. 1, Gohari Ardabili, A., Nemati, A. 2, & Latifi Darab, R. (2017). Modeling of strawberry drying process using infrared dryer by genetic algorithm–artificial neural network method. Journal Food Science and Technology, 14, 105–114.

Salehi, F., & Razavi, S. M. A. (2012). Dynamic modeling of flux and total hydraulic resistance in nanofiltration treatment of regeneration waste brine using artificial neural networks. Desalination and Water Treatment, 41(1–3), 95–104. https://doi.org/10.1080/19443994.2012.664683

Sawicka, B. (2020). Post-Harvest Losses of Agricultural Produce (pp. 654–669). https://doi.org/10.1007/978-3-319-95675-6_40

Seyedabadi, M. M., Aghajanzadeh, S. S., Kashaninejad, M., & Ziaiifar, A. M. (2017). INVESTIGATION OF THE EFFECT OF MICROWAVE ON SOME PHYSICOCHEMICAL PROPERTIES OF SOUR ORANGE JUICE.

Soleimanzadeh, B., Hemati, L., Yolmeh, M., & Salehi, F. (2015). GA-ANN and ANFIS models and salmonella enteritidis inactivation by ultrasound. Journal of Food Safety, 35(2), 220–226. https://doi.org/10.1111/jfs.12174

Taghadomi-Saberi, S., Mas Garcia, S., Allah Masoumi, A., Sadeghi, M., & Marco, S. (2018). Classification of Bitter Orange Essential Oils According to Fruit Ripening Stage by Untargeted Chemical Profiling and Machine Learning. Sensors, 18(6), 1922. https://doi.org/10.3390/s18061922

Tavar, M., Rabbani, H., Gholami, R., Ahmadi, E., & Kurtulmus, F. (2024). Investigating the Effect of Packaging Conditions on the Properties of Peeled Garlic by Using Artificial Neural Network (ANN). Packaging Technology and Science, 37(8), 755–767. https://doi.org/10.1002/pts.2819

Valentines, M. C., Vilaplana, R., Torres, R., Usall, J., & Larrigaudiere, C. (2005). Specific roles of enzymatic browning and lignification in apple disease resistance. Postharvest Biology and Technology, 36(3), 227–234.

Wypij, M., Trzcińska-Wencel, J., Golińska, P., Avila-Quezada, G. D., Ingle, A. P., & Rai, M. (2023). The strategic applications of natural polymer nanocomposites in food packaging and agriculture: Chances, challenges, and consumers’ perception. Frontiers in Chemistry, 10. https://doi.org/10.3389/fchem.2022.1106230

X, L., & W, W. (1998). Study on compressive properties of apple. Journal of Northwestern Agricultural University, 26(2), 107–108.

Yadav, N., & Kaur, R. (2024). Innovations in Packaging to Monitor and Maintain the Quality of the Food Products. Journal of Packaging Technology and Research, 8(1), 15–50. https://doi.org/10.1007/s41783-024-00163-4

Keywords

Physicochemical Properties

Banana

Neural Network

Subjects

Postharvest Technology