Analysis of Machine Learning Algorithms for Anomaly Detection on Edge Devices.

The Internet of Things (IoT) consists of small devices or a network of sensors, which permanently generate huge amounts of data. Usually, they have limited resources, either computing power or memory, which means that raw data are transferred to central systems or the cloud for analysis. Lately, the idea of moving intelligence to the IoT is becoming feasible, with machine learning (ML) moved to edge devices. The aim of this study is to provide an experimental analysis of processing a large imbalanced dataset (DS2OS), split into a training dataset (80%) and a test dataset (20%). The training dataset was reduced by randomly selecting a smaller number of samples to create new datasets Di (i = 1, 2, 5, 10, 15, 20, 40, 60, 80%). Afterwards, they were used with several machine learning algorithms to identify the size at which the performance metrics show saturation and classification results stop improving with an F1 score equal to 0.95 or higher, which happened at 20% of the training dataset. Further on, two solutions for the reduction of the number of samples to provide a balanced dataset are given. In the first, datasets DRi consist of all anomalous samples in seven classes and a reduced majority class ('NL') with i = 0.1, 0.2, 0.5, 1, 2, 5, 10, 15, 20 percent of randomly selected samples. In the second, datasets DCi are generated from the representative samples determined with clustering from the training dataset. All three dataset reduction methods showed comparable performance results. Further evaluation of training times and memory usage on Raspberry Pi 4 shows a possibility to run ML algorithms with limited sized datasets on edge devices.

Evaluation on home storage performance of table grape based on sensory quality and consumers' satisfaction.

With continuous rise of table grapes consumption and increased public awareness of food safety, the quality control of grapes in storage after purchase is not sufficiently examined. Home storage constitutes the last and important stage in grape supply chain. Literature review shows that few researches on grape quality focus on the home storage stage compared with numerous researches reported on the quality control during postharvest and transportation process. This paper reports the performance evaluation of grape quality at home storage and consumers' satisfaction using integrated sensory evaluations. The internal attributes, including Texture, Taste and Odor of the table grapes and the appearance indices, Color and Cleanliness are examined. Key results show that during home storage, all the internal attributes decrease rapidly as time goes on, and cleanliness and color appear to be deteriorating in a lower speed. A comprehensive quality index was created to measure the quality of table grape which has high correlation with the Overall acceptability perceived by consumers.

Assessment of 'Golden Delicious' Apples Using an Electronic Nose and Machine Learning to Determine Ripening Stages.

Consumers often face a lack of information regarding the quality of apples available in supermarkets. General appearance factors, such as color, mechanical damage, or microbial attack, influence consumer decisions on whether to purchase or reject the apples. Recently, devices known as electronic noses provide an easy-to-use and non-destructive assessment of ripening stages based on Volatile Organic Compounds (VOCs) emitted by the fruit. In this study, the 'Golden Delicious' apples, stored and monitored at the ambient temperature, were analyzed in the years 2022 and 2023 to collect data from four Metal Oxide Semiconductor (MOS) sensors (MQ3, MQ135, MQ136, and MQ138). Three ripening stages (less ripe, ripe, and overripe) were identified using Principal Component Analysis (PCA) and the K-means clustering approach from various datasets based on sensor measurements in four experiments. After applying the K-Nearest Neighbors (KNN) model, the results showed successful classification of apples for specific datasets, achieving an accuracy higher than 75%. For the dataset with measurements from all experiments, an impressive accuracy of 100% was achieved on specific test sets and on the evaluation set from new, completely independent experiments. Additionally, correlation and PCA analysis showed that choosing two or three sensors can provide equally successful results. Overall, the e-nose results highlight the importance of analyzing data from several experiments performed over a longer period after the harvest of apples. There are similarities and differences in investigated VOC parameters (ethylene, esters, alcohols, and aldehydes) for less or more mature apples analyzed during autumn or spring, which can improve the determination of the ripening stage with higher predicting success for apples investigated in the spring.