K-fold Cross-validation Research Articles

CT and MRI radiomics of bone and soft-tissue sarcomas: an updated systematic review of reproducibility and validation strategies.

To systematically review radiomic feature reproducibility and model validation strategies in recent studies dealing with CT and MRI radiomics of bone and soft-tissue sarcomas, thus updating a previous version of this review which included studies published up to 2020. A literature search was conducted on EMBASE and PubMed databases for papers published between January 2021 and March 2023. Data regarding radiomic feature reproducibility and model validation strategies were extracted and analyzed. Out of 201 identified papers, 55 were included. They dealt with radiomics of bone (n= 23) or soft-tissue (n = 32) tumors. Thirty-two (out of 54 employing manual or semiautomatic segmentation, 59%) studies included a feature reproducibility analysis. Reproducibility was assessed based on intra/interobserver segmentation variability in 30 (55%) and geometrical transformations of the region of interest in 2 (4%) studies. At least one machine learning validation technique was used for model development in 34 (62%) papers, and K-fold cross-validation was employed most frequently. A clinical validation of the model was reported in 38 (69%) papers. It was performed using a separate dataset from the primary institution (internal test) in 22 (40%), an independent dataset from another institution (external test) in 14 (25%) and both in 2 (4%) studies. Compared to papers published up to 2020, a clear improvement was noted with almost double publications reporting methodological aspects related to reproducibility and validation. Larger multicenter investigations including external clinical validation and the publication of databases in open-access repositories could further improve methodology and bring radiomics from a research area to the clinical stage. An improvement in feature reproducibility and model validation strategies has been shown in this updated systematic review on radiomics of bone and soft-tissue sarcomas, highlighting efforts to enhance methodology and bring radiomics from a research area to the clinical stage. • 2021-2023 radiomic studies on CT and MRI of musculoskeletal sarcomas were reviewed. • Feature reproducibility was assessed in more than half (59%) of the studies. • Model clinical validation was performed in 69% of the studies. • Internal (44%) and/or external (29%) test datasets were employed for clinical validation.

Efficiently improving the Wi-Fi-based human activity recognition, using auditory features, autoencoders, and fine-tuning

Human activity recognition (HAR) based on Wi-Fi signals has attracted significant attention due to its convenience and the availability of infrastructures and sensors. Channel State Information (CSI) measures how Wi-Fi signals propagate through the environment. However, many scenarios and applications have insufficient training data due to constraints such as cost, time, or resources. This poses a challenge for achieving high accuracy levels with machine learning techniques. In this study, multiple deep learning models for HAR were employed to achieve acceptable accuracy levels with much less training data than other methods. A pretrained encoder trained from a Multi-Input Multi-Output Autoencoder (MIMO AE) on Mel Frequency Cepstral Coefficients (MFCC) from a small subset of data samples was used for feature extraction. Then, fine-tuning was applied by adding the encoder as a fixed layer in the classifier, which was trained on a small fraction of the remaining data. The evaluation results (K-fold cross-validation and K = 5) showed that using only 30% of the training and validation data (equivalent to 24% of the total data), the accuracy was improved by 17.7% compared to the case where the encoder was not used (with an accuracy of 79.3% for the designed classifier, and an accuracy of 90.3% for the classifier with the fixed encoder). While by considering more calculational cost, achieving higher accuracy using the pretrained encoder as a trainable layer is possible (up to 2.4% improvement), this small gap demonstrated the effectiveness and efficiency of the proposed method for HAR using Wi-Fi signals.

SlumberNet: deep learning classification of sleep stages using residual neural networks

Sleep research is fundamental to understanding health and well-being, as proper sleep is essential for maintaining optimal physiological function. Here we present SlumberNet, a novel deep learning model based on residual network (ResNet) architecture, designed to classify sleep states in mice using electroencephalogram (EEG) and electromyogram (EMG) signals. Our model was trained and tested on data from mice undergoing baseline sleep, sleep deprivation, and recovery sleep, enabling it to handle a wide range of sleep conditions. Employing k-fold cross-validation and data augmentation techniques, SlumberNet achieved high levels of overall performance (accuracy = 97%; F1 score = 96%) in predicting sleep stages and showed robust performance even with a small and diverse training dataset. Comparison of SlumberNet's performance to manual sleep stage classification revealed a significant reduction in analysis time (~ 50 × faster), without sacrificing accuracy. Our study showcases the potential of deep learning to facilitate sleep research by providing a more efficient, accurate, and scalable method for sleep stage classification. Our work with SlumberNet further demonstrates the power of deep learning in mouse sleep research.

Integrating land use and traffic to spatial prediction of particulate matter

With degrading air quality and inefficient monitoring systems, spatial prediction of pollutants has become necessary. Air pollution exposure causes significant impacts on human health in urban areas around the world. The current study focuses on developing a deep learning model for spatial prediction of particulate matter at locations where the monitored data is unavailable. Delhi, India, is selected as the study area due to its high air pollution and fewer fixed monitoring stations throughout the city. The parameters used for the prediction modeling included PM2.5 from fixed monitoring stations, PM2.5 from mobile sensors, meteorological parameters (atmospheric temperature and relative humidity), traffic flow data, and land use parameters. The data is taken for six months, from November 2022 to April 2023, at an interval of 15 min. Four buffer sizes are used to find the spatial variability. The ANN model is used for the spatial predictions using the above-mentioned parameters. K-fold cross-validation is used to validate the robustness of the model. The SHAP summary plot is used to identify the feature importance of the features used in the model. It is observed that the integration of the land use data and traffic data increases the prediction capability of the model.

Manufacturing process encoding through natural language processing for prediction of material properties

Knowledge of manufacturing processes is crucial to determine the final properties of a material, thus this work focuses on analyzing the relationship between final properties, chemical composition, and manufacturing process through data analysis. Furthermore, techniques of natural language processing are used to encode the manufacturing process as input in the neural network. The work consisted of two main parts: firstly, the relevant data was gathered, cleaned-up, and analyzed using statistical and probabilistic methods, K-means, and Principal Components Analysis (PCA), and secondly, a model was developed to predict elongation, yield, and tensile strength. Fully Connected Neural Network (FCNN) algorithms were used to build the aforementioned model. In addition, in order to avoid overfitting and evaluate the model dropout function and K-fold cross-validation are incorporated. Results demonstrated reasonable accuracy in elongation, yield, and tensile strength. Two cases of mechanical properties prediction of stainless steel alloy were presented, first, an existing alloy that was not in the training and test set, and second a suggestion of a new stainless steel alloy, which combines good YS, UTS, and Pitting resistance equivalent number (PREN). Additionally, it was considered an example from the TRIP family to showcase the tool’s versatility across various steel types.

Enhancement in the safety and reliability of Pressurized Water reactors using Machine Learning approach

The increasing interest in artificial intelligence and automation within the nuclear industry stems from the hope of elevating the safety and reliability of nuclear power plants by minimizing the impact of human errors. This paper explores the capabilities of machine-learning based fault detection and diagnosis (FDD) models in accurately classifying transient events in a nuclear power plant using the Classification Learners Application from MATLAB. The scope of the research is narrowed to the identification of transient events of the same nature occurring at various locations within the plant. To accomplish this, the Generic Pressurized Water Reactor (GPWR) from the Western Services Cooperation (WSC) was used to simulate 14 transients in the primary system and 10 transients in the secondary system, resulting in two distinct datasets of 108,000 and 79,200 observations. Nine types of classifiers, including Decision Trees, Discriminant Analysis, Support Vector Machines, Logistic Regression, Nearest Neighbors, Naive Bayes, Kernel Approximation, Ensembles, and Neural Networks, with a total of 33 predefined models, were trained and validated. The K-Nearest Neighbors model achieved the highest accuracy of 97% in the primary system, while the Efficient Linear Discriminant and Logistic Regression models achieved the highest average accuracy of 99% in the secondary system. The assessment of the top-performing models was conducted through a comprehensive analysis employing key classification model evaluation metrics, such as the confusion matrix, accuracy, precision, recall, and the F1 score. In an effort to optimize the models, the study integrated feature selection using the minimum Redundancy Maximum Relevance (mRMR) algorithm. Additionally, the investigation explored different validation schemes, including both holdout and k-fold cross-validation schemes, resulting in a substantial reduction in training time without significantly compromising the overall accuracy of the models. The optimized models in this study demonstrated remarkable prediction accuracies, consistently exceeding 94%. Despite the complexity of the models and the intricacies involved in the training process, the training times ranged from 10 to 1800 s, reflecting the efficiency and robustness of the implemented optimization techniques. The results of this study underscore the potential of machine learning models in identifying transient events in nuclear reactors, showcasing their promising capabilities while maintaining low computational and execution costs. This suggests their efficacy in optimizing safe and efficient operational practices within nuclear facilities.

Twitter Data Sentimen Analysis 2024 Presidential Candidate Using Algorithm Naïve Bayes Classifier By Methods K-Fold Cross Validation

Indonesia implements a democratic system by involving the public in General Elections (Pemilu) for specific political positions. The active community expresses opinions on social media, especially regarding the 2024 Presidential Election (Pilpres) and respective presidential candidates, which have become trending topics on Twitter. The analysis used to absorb these tweets into information is sentimen analysis using the Naïve Bayes Classifier algorithm with the K-fold Cross-Validation method. Through the stages of pre-processing, weighting, labeling, classification using NBC, and testing using a Confusion Matrix, The results of the classification from NBC showed that Anies got 80% positive tweets and 20% negative tweets from 1186 tweets, Prabowo Subianto got 78% positive tweets and 22% negative tweets from 1149 tweets, and Ganjar Pranowo got 77% positive tweets and 23% negative tweets from 1075 tweets. Testing process was carried out using the NBC algorithm with the K-Fold Cross Validation method using values k=1 to k=10. The function of K-Fold Cross Validation is to maximize the confusion matrix result. It can be conclude that Anies Baswedan has the highest score in iteration 4, namely a precision value of 90%, a recall value of 99%, and the accurary value of 91%. Furthemore, Ganjar Pranowo had the highest score in iteration 9, namely a precision value of 95%,a recall value of 97%, and an accuracy value of 92%. Meanwhile, Prabowo Subianto had the highest score in iteration 9, namely a precision value of 97%, a recall value of 99%, and an accuracy value of 94%.

Deep neural networks can differentiate thyroid pathologies on infrared hyperspectral images

Background and objectiveThe thyroid is a gland responsible for producing important body hormones. Several pathologies can affect this gland, such as thyroiditis, hypothyroidism, and thyroid cancer. The visual histological analysis of thyroid specimens is a valuable process that enables pathologists to detect diseases with high efficiency, providing the patient with a better prognosis. Existing computer vision systems developed to aid in the analysis of histological samples have limitations in distinguishing pathologies with similar characteristics or samples containing multiple diseases. To overcome this challenge, hyperspectral images are being studied to represent biological samples based on their molecular interaction with light. MethodsIn this study, we address the acquisition of infrared absorbance spectra from each voxel of histological specimens. This data is then used for the development of a multiclass fully-connected neural network model that discriminates spectral patterns, enabling the classification of voxels as healthy, cancerous, or goiter. ResultsThrough experiments using the k-fold cross-validation protocol, we obtained an average accuracy of 93.66 %, a sensitivity of 93.47 %, and a specificity of 96.93 %. Our results demonstrate the feasibility of using infrared hyperspectral imaging to characterize healthy tissue and thyroid pathologies using absorbance measurements. The proposed deep learning model has the potential to improve diagnostic efficiency and enhance patient outcomes.

Deep Learning for Generalized EEG Seizure Detection after Hypoxia-Ischemia-Preclinical Validation.

Brain maturity and many clinical treatments such as therapeutic hypothermia (TH) can significantly influence the morphology of neonatal EEG seizures after hypoxia-ischemia (HI), and so there is a need for generalized automatic seizure identification. This study validates efficacy of advanced deep-learning pattern classifiers based on a convolutional neural network (CNN) for seizure detection after HI in fetal sheep and determines the effects of maturation and brain cooling on their accuracy. The cohorts included HI-normothermia term (n = 7), HI-hypothermia term (n = 14), sham-normothermia term (n = 5), and HI-normothermia preterm (n = 14) groups, with a total of >17,300 h of recordings. Algorithms were trained and tested using leave-one-out cross-validation and k-fold cross-validation approaches. The accuracy of the term-trained seizure detectors was consistently excellent for HI-normothermia preterm data (accuracy = 99.5%, area under curve (AUC) = 99.2%). Conversely, when the HI-normothermia preterm data were used in training, the performance on HI-normothermia term and HI-hypothermia term data fell (accuracy = 98.6%, AUC = 96.5% and accuracy = 96.9%, AUC = 89.6%, respectively). Findings suggest that HI-normothermia preterm seizures do not contain all the spectral features seen at term. Nevertheless, an average 5-fold cross-validated accuracy of 99.7% (AUC = 99.4%) was achieved from all seizure detectors. This significant advancement highlights the reliability of the proposed deep-learning algorithms in identifying clinically translatable post-HI stereotypic seizures in 256Hz recordings, regardless of maturity and with minimal impact from hypothermia.

Suboptimal capability of individual machine learning algorithms in modeling small-scale imbalanced clinical data of local hospital.

In recent years, artificial intelligence (AI) has shown promising applications in various scientific domains, including biochemical analysis research. However, the effectiveness of AI in modeling small-scale, imbalanced datasets remains an open question in such fields. This study explores the capabilities of eight basic AI algorithms, including ridge regression, logistic regression, random forest regression, and others, in modeling a small, imbalanced clinical dataset (total n = 387, class 0 = 27, class 1 = 360) related to the records of the biochemical blood tests from the patients with multiple wasp stings (MWS). Through rigorous evaluation using k-fold cross-validation and comprehensive scoring, we found that none of the models could effectively model the data. Even after fine-tuning the hyperparameters of the best-performing models, the results remained below acceptable thresholds. The study highlights the challenges of applying AI to small-scale datasets with imbalanced groups in biochemical or clinical research and emphasizes the need for novel algorithms tailored to small-scale data. The findings also call for further exploration into techniques such as transfer learning and data augmentation, and they underline the importance of understanding the minimum dataset scale required for effective AI modeling in biochemical contexts.

A comparative analysis of boosting algorithms for chronic liver disease prediction

Chronic liver disease (CLD) is a major health concern for millions of people all over the globe. Early prediction and identification are critical for taking appropriate action at the earliest stages of the disease. Implementing machine learning methods in predicting CLD can greatly improve medical outcomes, reduce the burden of the condition, and promote proactive and preventive healthcare practices for those at risk. However, traditional machine learning has some limitations which can be mitigated through ensemble learning. Boosting is the most advantageous ensemble learning approach. This study aims to improve the performance of the available boosting techniques for CLD prediction. Seven popular boosting algorithms of Gradient Boosting (GB), AdaBoost, LogitBoost, SGBoost, XGBoost, LightGBM, and CatBoost, and two publicly available popular CLD datasets (Liver disease patient dataset (LDPD) and Indian liver disease patient dataset (ILPD)) of dissimilar size and demography are considered in this study. The features of the datasets are ascertained by exploratory data analysis. Additionally, hyperparameter tuning, normalisation, and upsampling are used for predictive analytics. The proportional importance of every feature contributing to CLD for every algorithm is assessed. Each algorithm's performance on both datasets is assessed using k-fold cross-validation, twelve metrics, and runtime. Among the five boosting algorithms, GB emerged as the best overall performer for both datasets. It attained 98.80% and 98.29% accuracy rates for LDPD and ILPD, respectively. GB also outperformed other boosting algorithms regarding other performance metrics except runtime.

Intelligent Classifiers for Football Player Performance Based on Machine Learning Models

The remarkable effectiveness of Machine Learning (ML) methodologies has led to a significant increase in their application across various academic domains, particularly in diverse sports sectors. Over the past decade, scholars have utilized Machine Learning (ML) algorithms in football for varied objectives, encompassing the analysis of football players' performances, injury prediction, market value forecasting, and action recognition. Nevertheless, there has been a scarcity of research addressing the evaluation of football players' performance, which is a noteworthy concern for coaches. Hence, the objective of this work is to categorize the performance of football players into active, normal, or weak based on activity features. This will be achieved through the utilization of the Performance Evaluation Machine Learning Model (PEMLM), employing two novel datasets that cover both training and match sessions. To attain this goal, seven machine learning methods are applied, namely Random Forest, Decision Tree, Logistic Regression, Support Vector Machine, Gaussian Naïve Bayes, Multi-Layer Perceptron, and K-Nearest Neighbor. The findings indicate that in the dataset corresponding to match sessions, the Decision Tree classifier attains the highest accuracy (100%) and the shortest test time. In contrast, the K-Nearest Neighbor demonstrates the best accuracy (96%) and a reasonable test time for the training dataset. These reported metrics underscore the reliability and validity of the proposed assessment approach in evaluating the performance of football players in online games. The results are verified and the models are assessed for overfitting through a k-fold cross-validation process.

Data-driven crack image-based seismic failure mode identification for damaged RC columns

The seismic failure mode identification of reinforced concrete columns is crucial for post-earthquake stability and safety assessment of the damaged concrete buildings. In this paper, a multi-feature crack image-based methodology is proposed for non-destructive non-contact failure mode prediction in reinforced concrete columns damaged in an earthquake. A databank of 455 surface crack patterns from quasi-static experiments on 111 reinforced concrete columns with a variety of structural and geometric properties is collected by the authors aimed at the development and validation of the methodology. For the intricacy retrieval of the surface crack maps, multi-feature fractal-based procedures are taken into consideration. Various machine learning algorithms are employed to train the seismic failure mode predictive models using the characteristics of the damaged RC columns. Ten scenarios are defined based on the image-extracted indices, and structural and geometric characteristics of rectangular reinforced concrete columns. Hyperparameter tuning of the machine learning algorithms using GridsearchCV function has optimized the performance of the models. A k-fold cross-validation procedure is also executed to examine the robustness of the results. Results reveal that the predictive model in which the image-extracted indices and structural parameters are jointly used as input features of supervised learning provides 98% accuracy for the testing dataset with Extreme gradient boost algorithm. In addition, among the scenarios that only image-derived parameters are considered as input features, an accuracy of 83% is obtained with Extreme gradient boost algorithm for the testing dataset.

Convolutional Neural Networks for Raw Signal Classification in CNC Turning Process Monitoring.

This study addresses the need for advanced machine learning-based process monitoring in smart manufacturing. A methodology is developed for near-real-time part quality prediction based on process-related data obtained from a CNC turning center. Instead of the manual feature extraction methods typically employed in signal processing, a novel one-dimensional convolutional architecture allows the trained model to autonomously extract pertinent features directly from the raw signals. Several signal channels are utilized, including vibrations, motor speeds, and motor torques. Three quality indicators-average roughness, peak-to-valley roughness, and diameter deviation-are monitored using a single model, resulting in a compact and efficient classifier. Training data are obtained via a small number of experiments designed to induce variability in the quality metrics by varying feed, cutting speed, and depth of cut. A sliding window technique augments the dataset and allows the model to seamlessly operate over the entire process. This is further facilitated by the model's ability to distinguish between cutting and non-cutting phases. The base model is evaluated via k-fold cross validation and achieves average F1 scores above 0.97 for all outputs. Consistent performance is exhibited by additional instances trained under various combinations of design parameters, validating the robustness of the proposed methodology.

Robust human locomotion and localization activity recognition over multisensory.

Human activity recognition (HAR) plays a pivotal role in various domains, including healthcare, sports, robotics, and security. With the growing popularity of wearable devices, particularly Inertial Measurement Units (IMUs) and Ambient sensors, researchers and engineers have sought to take advantage of these advances to accurately and efficiently detect and classify human activities. This research paper presents an advanced methodology for human activity and localization recognition, utilizing smartphone IMU, Ambient, GPS, and Audio sensor data from two public benchmark datasets: the Opportunity dataset and the Extrasensory dataset. The Opportunity dataset was collected from 12 subjects participating in a range of daily activities, and it captures data from various body-worn and object-associated sensors. The Extrasensory dataset features data from 60 participants, including thousands of data samples from smartphone and smartwatch sensors, labeled with a wide array of human activities. Our study incorporates novel feature extraction techniques for signal, GPS, and audio sensor data. Specifically, for localization, GPS, audio, and IMU sensors are utilized, while IMU and Ambient sensors are employed for locomotion activity recognition. To achieve accurate activity classification, state-of-the-art deep learning techniques, such as convolutional neural networks (CNN) and long short-term memory (LSTM), have been explored. For indoor/outdoor activities, CNNs are applied, while LSTMs are utilized for locomotion activity recognition. The proposed system has been evaluated using the k-fold cross-validation method, achieving accuracy rates of 97% and 89% for locomotion activity over the Opportunity and Extrasensory datasets, respectively, and 96% for indoor/outdoor activity over the Extrasensory dataset. These results highlight the efficiency of our methodology in accurately detecting various human activities, showing its potential for real-world applications. Moreover, the research paper introduces a hybrid system that combines machine learning and deep learning features, enhancing activity recognition performance by leveraging the strengths of both approaches.

Utilizing machine learning for flow zone indicators prediction and hydraulic flow unit classification

Reservoir characterization, essential for understanding subsurface heterogeneity, often faces challenges due to scale-dependent variations. This study addresses this issue by utilizing hydraulic flow unit (HFU) zonation to group rocks with similar petrophysical and flow characteristics. Flow Zone Indicator (FZI), a crucial measure derived from pore throat size, permeability, and porosity, serves as a key parameter, but its determination is time-consuming and expensive. The objective is to employ supervised and unsupervised machine learning to predict FZI and classify the reservoir into distinct HFUs. Unsupervised learning using K-means clustering and supervised algorithms including Random Forest (RF), Extreme Gradient Boosting (XGB), Support Vector Machines (SVM), and Artificial Neural Networks (ANN) were employed. FZI values from RCAL data formed the basis for model training and testing, then the developed models were used to predict FZI in unsampled locations. A methodical approach involves 3 k-fold cross-validation and hyper-parameter tuning, utilizing the random search cross-validation technique over 50 iterations was applied to optimize each model. The four applied algorithms indicate high performance with coefficients determination (R2) of 0.89 and 0.91 in training and testing datasets, respectively. RF showed the heist performance with training and testing R2 values of 0.957 and 0.908, respectively. Elbow analysis guided the successful clustering of 212 data points into 10 HFUs using k-means clustering and Gaussian mixture techniques. The high-quality reservoir zone was successfully unlocked using the unsupervised technique. It has been discovered that the areas between 2370–2380 feet and 2463–2466 feet are predicted to be high-quality reservoir potential areas, with average FZI values of 500 and 800, consecutively. The application of machine learning in reservoir characterization is deemed highly valuable, offering rapid, cost-effective, and precise results, revolutionizing decision-making in field development compared to conventional methods.

Predicting multi-horizon currency exchange rates using a stacked ensemble of random forest and SVR

A thorough exploration of the effects of a given minute’s currency exchange rates on subsequent 1, 5, 10, 15, 30, 45, and 60 minutes’ currency exchange rates is presented in this article, with machine learning and ensemble methods being applied. The focus is on twelve currency pairs, including EUR/AUD, EUR/GBP, and EUR/PLN, with a data set of per-minute logs of these pairs’ exchange rates from 2022 being leveraged. A stacked ensemble of Random Forest and Support Vector Regression (SVR) is used to predict future exchange rates. A comparison of this model is also made with the single RF, single SVR, and an average ensemble of RF and SVR models. The research method is further fortified by the use of k-fold cross-validation and ANOVA tests. The findings of the study reveal significant predictive accuracy of the stacked ensemble model, emphasizing the intricate interconnections of currency exchange rates. The potential of machine learning and ensemble techniques in predicting short-term currency exchange rates is underlined, thereby augmenting financial forecasting research.

Spatial Prediction of Soil Contaminants Using a Hybrid Random Forest–Ordinary Kriging Model

The accurate prediction of soil contamination in abandoned mining areas is necessary to address their environmental risks. This study employed a combined model of machine learning and geostatistics to predict the spatial distribution of soil contamination using heavy metal data collected in an abandoned metal mine. An exploratory data analysis was used to identify patterns in the collected data, the root mean squared error (RMSE) and coefficient of determination (R2) were used to verify the predicted values, and the model was validated using K-fold cross-validation. The prediction results were produced as a map by applying hyperparameter tuning to Random Forest (RF) and Ordinary Kriging (OK) through GridSearchCV using optimal parameter selections. Furthermore, the prediction residuals of the RF model were calculated, and the RF prediction map and OK interpolation results of the RF prediction residuals were summarized to construct an RF–OK prediction map. The RMSE and R2 values for the RF, OK, and RF–OK interpolation models were 66.214, 65.101, and 52.884 mg/kg and 0.867, 0.871, and 0.915, respectively. In addition, the optimization results with the minimum RMSE and maximum R2 were obtained through hyperparameter tuning. The proposed RF–OK hybrid model demonstrated superior prediction performance compared to the individual models.

Design of a traceability system for a coffee supply chain based on blockchain and machine learning

Purpose: This paper aims to develop a coffee supply chain traceability system based on Blockchain (BC) and Machine Learning (ML) with the aim of ensuring the quality of coffee beans production. BC functions to ensure supply chain performance, while the ML model ensures product quality.Design/methodology/approach: Smart Contracts will be built on the Ethereum Virtual Machine BC network based on Ethereum. The ML model to identify good and bad green coffee beans will be built using different YOLO algorithms, which will go through training and validation stages, namely using the k-fold cross validation method. The ML model algorithm is based on Convolutional Neural Network (CNN) using YOLOv5m, YOLOv6m and YOLOv7. The best model will be chosen based on the results of cross-validation with test data in the form of coffee image data that the model has never seen (unseen data). The whole process of building the ML model is done on the Google Collab Pro+ Virtual Machine.Findings: YOLOv5m outperformed the other models in both non-augmented and augmented training datasets, highlighting the proficiency of YOLOv5m in managing compact datasets and its resilience in the face of data augmentation, positioning it as a prime selection for quality discernment tasks within the realm of green coffee beans. The smart contracts offer an all-encompassing approach for user management, monitoring product status, and presenting traceability data within the framework of coffee plantation administration.Originality/value: This research contributes to the development of blockchain network as a solution to implement traceability systems along coffee supply chains in Indonesia. Moreover, it shows that while blockchain can ensure the process of production along the coffee supply to follow certain guidelines, machine learning can verify whether the product that was produced by utilizing BC is of high quality/acceptable.

Real-Time Tissue Classification Using a Novel Optical Needle Probe for Biopsy.

Core needle biopsy is a part of the histopathological process, which is required for cancerous tissue examination. The most common method to guide the needle inside of the body is ultrasound screening, which in greater part is also the only guidance method. Ultrasound screening requires user experience. Furthermore, patient involuntary movements such as breathing might introduce artifacts and blur the screen. Optically enhanced core needle biopsy probe could potentially aid interventional radiologists during this procedure, providing real-time information on tissue properties close to the needle tip, while it is advancing inside of the body. In this study, we used diffuse optical spectroscopy in a custom-made core needle probe for real-time tissue classification. Our aim was to provide initial characteristics of the smart needle probe in the differentiation of tissues and validate the basic purpose of the probe of informing about breaking into a desired organ. We collected optical spectra from rat blood, fat, heart, kidney, liver, lungs, and muscle tissues. Gathered data were analyzed for feature extraction and evaluation of two machine learning-based classifiers: support vector machine and k-nearest neighbors. Their performances on training data were compared using subject-independent k-fold cross-validation. The best classifier model was chosen and its feasibility for real-time automated tissue recognition and classification was then evaluated. The final model reached nearly 80% of correct real-time classification of rat organs when using the needle probe during real-time classification.