Machine Learning Approach Research Articles

Image analysis for classifying coffee bean quality using a multi feature and machine learning approach

Price and customer satisfaction depend on coffee bean quality. The coffee industry must analyze coffee bean quality. Global demand for robusta coffee is high. Coffee industry professionals mostly understand coffee bean quality. Thus, an image analysis using a computer vision-based approach for classifying robusta coffee bean quality is required. Image acquisition, region of interest (ROI) detection, pre-processing, segmentation, feature extraction, feature selection, and classification are covered in this study. A multi-feature derived based on color, shape, and texture features was employed in feature extraction, followed by feature selection using principal component analysis (PCA). Several machine-learning methods classified the coffee beans. The method performance was assessed using precision, recall, and accuracy. The selected features using the backpropagation neural network (BPNN) classifier outperformed others with 98.54% accuracy.

Patterns of Gene Expression Profiles Associated with Colorectal Cancer in Colorectal Mucosa by Using Machine Learning Methods.

Colorectal cancer (CRC) has a very high incidence and lethality rate and is one of the most dangerous cancer types. Timely diagnosis can effectively reduce the incidence of colorectal cancer. Changes in para-cancerous tissues may serve as an early signal for tumorigenesis. Comparison of the differences in gene expression between para-cancerous and normal mucosa can help in the diagnosis of CRC and understanding the mechanisms of development. This study aimed to identify specific genes at the level of gene expression, which are expressed in normal mucosa and may be predictive of CRC risk. A machine learning approach was used to analyze transcriptomic data in 459 samples of normal colonic mucosal tissue from 322 CRC cases and 137 non-CRC, in which each sample contained 28,706 gene expression levels. The genes were ranked using four ranking methods based on importance estimation (LASSO, LightGBM, MCFS, mRMR, and RF) and four classification algorithms (decision tree [DT], K-nearest neighbor [KNN], random forest [RF], and support vector machine [SVM]) were combined with incremental feature selection [IFS] methods to construct a prediction model with excellent performance. The top-ranked genes, namely, HOXD12, CDH1, and S100A12, were associated with tumorigenesis based on previous studies. This study summarized four sets of quantitative classification rules based on the DT algorithm, providing clues for understanding the microenvironmental changes caused by CRC. According to the rules, the effect of CRC on normal mucosa can be determined.

Spatial and spatiotemporal modelling of intra-urban ultrafine particles: A comparison of linear, nonlinear, regularized, and machine learning methods

BackgroundMachine learning methods are proposed to improve the predictions of ambient air pollution, yet few studies have compared ultrafine particles (UFP) models across a broad range of statistical and machine learning approaches, and only one compared spatiotemporal models. Most reported marginal differences between methods. This limits our ability to draw conclusions about the best methods to model ambient UFPs. ObjectiveTo compare the performance and predictions of statistical and machine learning methods used to model spatial and spatiotemporal ambient UFPs. MethodsDaily and annual models were developed from UFP measurements from a year-long mobile monitoring campaign in Quebec City, Canada, combined with 262 geospatial and six meteorological predictors. Various road segment lengths were considered (100/300/500 m) for UFP data aggregation. Four statistical methods included linear, non-linear, and regularized regressions, whereas eight machine learning regressions utilized tree-based, neural networks, support vector, and kernel ridge algorithms. Nested cross-validation was used for model training, hyperparameter tuning and performance evaluation. ResultsMean annual UFP concentrations was 13,335 particles/cm3. Machine learning outperformed statistical methods in predicting UFPs. Tree-based methods performed best across temporal scales and segment lengths, with XGBoost producing the overall best performing models (annual R2 = 0.78–0.86, RMSE = 2163–2169 particles/cm3; daily R2 = 0.47–0.48, RMSE = 8651–11,422 particles/cm3). With 100 m segments, other annual models performed similarly well, but their prediction surfaces of annual mean UFP concentrations showed signs of overfitting. Spatial aggregation of monitoring data significantly impacted model performance. Longer segments yielded lower RMSE in all daily models and for annual statistical models, but not for annual machine learning models. ConclusionsThe use of tree-based methods significantly improved spatiotemporal predictions of UFP concentrations, and to a lesser extent annual concentrations. Segment length and hyperparameter tuning had notable impacts on model performance and should be considered in future studies.

Non-contact power system fault diagnosis: a machine learning approach with electromagnetic current sensing

Modern power system protection schemes incorporate artificial intelligence (AI) techniques. However, in a conventional way, most of these schemes rely on the data of current and voltage collected from current transformer (CT) and potential transformer (PT) respectively. CTs suffer from the drawback of core saturation and impact the accuracy and effectiveness of intelligent methods. Also, it has the constraints of size, safety, and economy. The research here explores the effectiveness of magnetic sensors in advanced power system protection schemes as an alternative to traditional current sensing. In the presented work, a novel dataset is prepared by transforming transmission line currents into magnetic field components. Several supervised as well as unsupervised machine learning algorithms have been applied to this data instead of traditional currents and voltage for fault prediction. The paper discusses the comparative evaluation of these algorithms based on various performance metrices which reveals that Gaussian Naïve Bayes (GNB), K-nearest neighbor (KNN), random forest (RF), and extreme gradient boost (XGB) algorithms excel in fault detection, while multilayer perceptron (MLP) and KNN performs better fault classification. The findings promise the potential for developing compact, safe, and cost-effective protection schemes utilizing magnetic field sensors.

Covid-19 forecasting model based on machine learning approaches: a review

Covid-19 forecasting model based on machine learning approaches: a review

Regime Type and Data Manipulation: Evidence from the COVID-19 Pandemic

Context: This study examines whether autocratic governments are more likely than democratic governments to manipulate health data. The COVID-19 pandemic presents a unique opportunity for examining this question because of its global impact. Methods: Three distinct indicators of COVID-19 data manipulation were constructed for nearly all sovereign states. Each indicator was then regressed on democracy and controls for unintended misreporting. A machine learning approach was then used to determine whether any of the specific features of democracy are more predictive of manipulation. Findings: Democracy was found to be negatively associated with all three measures of manipulation, even after running a battery of robustness checks. Absence of opposition party autonomy and free and fair elections were found to be the most important predictors of deliberate undercounting. Conclusions: The manipulation of data in autocracies denies citizens the opportunity to protect themselves against health risks, hinders the ability of international organizations and donors to identify effective policies, and makes it difficult for scholars to assess the impact of political institutions on population health. These findings suggest that health advocates and scholars should use alternative methods to estimate health outcomes in countries where opposition parties lack autonomy or must participate in uncompetitive elections.

IoT-HGDS: Internet of Things integrated machine learning based hazardous gases detection system for smart kitchen

This paper proposes an Internet of Things (IoT) and Machine Learning (ML) integrated Hazardous Gas Detection System (IoT-HGDS) for smart kitchens. The design incorporates six tin-oxide-based metal–oxide–semiconductor (MOS) gas sensor arrays and one DHT22 (temperature & humidity sensor). This IoT-HGDS can detect different hazardous gases, Volatile Organic Compounds (VOCs), and odors responses released from the kitchen’s materials and transmit them to a Remote Data Processing Centre (RDPC) through Amazon-Web Services (AWS) in real time. In this experiment, we collected 150×9=1350 samples from 9 kitchen materials like ghee, milk, liquid petroleum gas (LPG), bread, mustard oil, compressed natural gas (CNG), pigeon peas, refined oil, and kerosene. The Standardized Independent Component Analysis (SICA) pre-processing technique has been used to clean data, standardise the features, and remove outliers. ML approaches like Logistic Regression (LR), Adaptive Boosting (AdaBoost) and Regularized Discriminant Analysis (RDA) have been applied for accurate identification of gases/VOCs class and provide immediate alerts to improve kitchen safety. The SICA-RDA classifier outperformed (highest accuracy at 97.78 %) as compared to LR and AdaBoost in terms of performance and balanced precision, recall, and F1-Score. LR has the lowest performance in all metrics. LPG has the lowest Mean Squared Error (MSE) of 6.62×10−7, while CNG has the highest MSE of 3.60×10−4. This system can intelligently preserve gases, ensure safety precautions, and prevent accidents in the kitchens.

Evaluating the electronic and structural basis of carbon selenide-based quantum dots as photovoltaic design materials: A DFT and ML analysis

We present a new study on the design, discovery and space generation of carbon selenide based photovoltaic (PV) materials. By extending acceptors and leveraging density functional theory (DFT) and machine learning (ML) analysis, we discover new QDs with remarkable PV properties. We employ various ML models, to correlate the exciton binding energy (Eb) of 938 relevant compounds from literature with their molecular descriptors of structural features that influence their performance. Our study demonstrates the potential of ML approaches in streamlining the design and discovery of high-efficiency PV materials. Also the RDKit computed molecular descriptors correlates with PV parameters revealed maximum absorption (λmax) ranges of 509–531 nm, light harvesting efficiency (LHE) above 92 %, Open Circuit Voltage (Voc) of 0.22–0.45 V, and short Circuit (Jsc) currents of 37.92–42.75 mA/cm2. Their Predicted Power Conversion Efficiencies (PCE) using the Scharber method reaches upto 09–13 %. This study can pave the way for molecular descriptor-based design of new PV materials, promising a paradigm shift in the development of high-efficiency solar energy conversion technologies.

Autoencoder-based drug synergy framework for malignant diseases

Drug combination emerges as a viable option for the treatment of malignant diseases. Drug combination outperforms monotherapy by improving therapeutic efficacy, reducing toxicity, and overcoming drug resistance. To find viable drug combinations it is difficult to traverse empirically because of enormous combinational space. Machine learning and deep learning approaches are used to uncover novel synergistic drug combinations in enormous combinational space. Here, AESyn, a novel autoencoder-based drug synergy framework for malignant diseases using a bag of words encoding is proposed. The bag of word encoding technique is used to extract drug-targeted genes. The framework utilized screening data from NCI-ALMANAC, and O’Neil datasets. Autoencoders take drug embeddings with drug-targeted genes as input for processing. The autoencoder in the proposed framework is used to extract drug features. The proposed framework is evaluated on classification and regression metrics. The performance of the proposed framework is compared with existing methods of drug synergy. According to the findings, the proposed framework achieved high performance with an accuracy of 95%, AUROC of 94.2%, and MAPE of 7.2. The autoencoder-based framework for malignant diseases using an encoding technique provides a stable, order-independent drug synergy prediction.

Comprehensive Performance Evaluation of Machine Learning Algorithms for detecting DDoS attacks in SDN

<div align="center">Software-Defined Networking (SDN) revolutionizes networking by separating control logic and data forwarding, enhancing security against threats like Distributed Denial of Service (DDoS) attacks. These attacks flood control plane bandwidth, causing SDN network failures. Recent studies emphasize the efficacy of machine learning and statistical approaches in identifying and mitigating these security risks. However, there has been a lack of focus on employing ensembling techniques, amalgamating diverse machine learning models, selecting pertinent features, and utilizing oversampling techniques to balance categorical data. Our study evaluates 20 machine-learning models, emphasizing feature engineering and addressing class imbalance using Synthetic Minority Oversampling TEchnique (SMOTE). The results indicate that Ensemble methods such as LGBM Classifier, Random Forest Classifier, XGB Classifier, Decision Tree Classifier attained near-perfect scores (almost 100%) across all metrics, suggesting potential overfitting. Conversely, models like AdaBoost Classifier, K-Neighbors Classifier, and SVC exhibited slightly lower (99%) but realistic performance, underscoring the intricacy of accurate prediction in cybersecurity. Simpler models, including Logistic Regression, Linear Discriminant Analysis, and Gaussian Naive Bayes, demonstrated moderate to low accuracy, approximately around 70%. These findings stress the imperative need for a nuanced approach in the selection and fine-tuning of machine learning models to ensure effective DDoS detection in SDN environments.</div>

ML enhanced measurement of the electrostatic charge distribution of powder conveyed through a duct

The electrostatic charge acquired by powders during transport through ducts can cause devastating dust explosions. Our recently developed laser-optical measurement technique can resolve the powder charge along a one-dimensional (1D) path. However, the charge across the duct’s complete two-dimensional (2D) cross-section, which is the critical parameter for process safety, is generally unavailable due to limited optical access. To estimate the complete powder charge distribution in a conveying duct, we propose a machine learning (ML) approach using a shallow neural network (SNN). The ML algorithm is trained with cross-sectional data extracted from four different three-dimensional direct numerical simulations of a turbulent duct flow with varying particle size. Through this training with simulation data, the ML algorithm can estimate the powder charge distribution in the duct’s cross-section based on only 1D measurements. The results reveal an average L1-error of the reconstructed 2D cross-section of 1.63%.

Geochemical fingerprints, evolution, and driving forces of groundwater in an alpine basin on Tibetan Plateau: Insights from unsupervised machine learning and objective weight allocation approaches

Study RegionTongde Basin, a typical alpine basin on the northeastern Tibetan Plateau Study focusResearch on the hydrochemical pattern and availability of groundwater is very limited in alpine regions, which greatly hinders the rational ulitilization and scientific management of groundwater resources. Multiple approaches integrating self-organizing map, K-means, entropy weight were employed to elucidate the distinct hydrochemical evolution patterns and controlling mechanisms of groundwater in alpine basins with dense human activities. New hydrological insights for the regionMultiple hydrochemical evolution patterns were discovered for groundwater in present alpine basin on northeastern Tibetan Plateau. Groundwater presents a gradual hydrochemical evolution from the piedmont to the riverside under the natural regulation with the increasing of TDS and major ions along the flow path. While, groundwater at some sporadic sites is featured with relatively high total alkalinity due to carbonate mineral dissolution. The riverine groundwaters are regulated by the strong hydrodynamic condition of river and demonstrate fresher hydrochemical characteristics. Groundwaters in the irrigation areas are featured by elevated nitrate levels with a maximum of 42.86mg/L, attributed to agricultural pollution inputs. A conceptual model was established to illustrate the distinct hydrochemical fingerprints, evolution patterns and controlling mechanisims of groundwater in alpine basins that with dense human activities. This research can enhance the understanding of groundwater chemical status, genesis and availability, and is beneficial for the conservation and sustainable management of groundwater resources in ecologically fragile alpine regions worldwide.

Sarcasm detection on social data: heuristic search and deep learning

<span lang="EN-US">Due to the significant surge in online activity, sarcasm detection (SD) has attracted major attention in social media networks. Sarcasm is a lexical item of negative sentiments or dislikes by utilizing exaggerated language constructs. SD has created a natural language processing (NLP) procedure focused on the intricate and unclear aspects of sarcasm, primarily used in sentiment analysis (SA), human-computer interaction, and various NLP applications. Concurrently, advancements in machine learning (ML) approaches facilitate the creation of effective SD systems. This manuscript presents the future search algorithm with deep learning assisted sarcasm detection and classification on social networking data (FSADL-SDCSND) approach. The major intent of the FSADL-SDCSND approach is in the effective and automated recognition of sarcastic text. In the presented FSADL-SDCSND technique, several data pre-processing stages are achieved to transform the data into a compatible format. Besides, the FSADL-SDCSND approach applies a bidirectional serial-parallel long short-term memory (BS-PLSTM) approach for SD and classification. The hyperparameter tuning process is accomplished by employing the future search algorithm for improving the recognition of the BS-PLSTM model. For superior output of the FSADL-SDCSND model, a sequence of simulations can be applied. The investigational outputs highlighted the improved solutions of the FSADL-SDCSND model with other approaches under diverse performance measures.</span>

A terrain data collection sensor box towards a better analysis of terrains conditions

<p>Autonomous mobile robots are increasingly used across various applications, relying on multiple sensors for environmental awareness and efficient task execution. Given the unpredictability of human environments, versatility is crucial for these robots. Their performance is largely determined by how they perceive their surroundings. This paper introduces a machine learning (ML) approach focusing on land conditions to enhance a robot’s locomotion. The authors propose a method to classify terrains for data collection, involving the design of an apparatus to gather field data. This design is validated by correlating collected data with the output of a standard ML model for terrain classification. Experiments show that the data from this apparatus improves the accuracy of the ML classifier, highlighting the importance of including such data in the dataset.</p>

Leveraging transformer models for enhanced temperature forecasting: a comparative analysis in the Beni Mellal region

The remarkable impact of transformers in artificial intelligence, exemplified by applications like GPT-3 in language processing, has sparked interest in their potential for time series analysis. This study aims to explore whether transformers, specifically temporal fusion transformers (TFT), can outperform conventional methods in this domain. The research question is whether TFT exhibits superior performance compared to conventional recurrent neural network (RNN) methods, specifically gated recurrent unit (GRU), and traditional machine learning approaches, notably autoregressive integrated moving average (ARIMA), in the context of time series analysis and temperature prediction. A comparative analysis is conducted among three models: ARIMA, GRU, and TFT. The study utilizes time series data spanning from 1984 to the end of 2022. The models’ performances are evaluated using multiple metrics: mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), and the coefficient of determination (R2). The TFT model achieves the lowest MAE, indicating high accuracy in its predictions. It outperforms both the RNN and traditional machine learning in temperature prediction tasks. Integrating the TFT model with the FAO penman-monteith method could improve irrigation scheduling due to more accurate temperature predictions, potentially enhancing water efficiency and crop yields.

Machine learning approach for predicting high JSC donor molecules in fullerene-typed organic solar cells

This study aimed to identify donor molecules that enhance the JSC in fullerene-type organic thin-film solar cells using materials informatics. After performing principal component analysis and Random Forest for feature selection, LASSO and Ridge regressions and SVR were developed. A genetic algorithm generated 250 new donor molecules, and SVR predicted that (i) 4H-cyclopentadithiophene, (ii) fluorine-containing structures, and (iii) C = O groups adjacent to thiophenes improve the JSC.

Data-driven model assessment: A comparative study for ship response determination

Several machine learning approaches to determine ship responses via data-driven models have been applied. Input features and parameters used relied on time-series analyses obtained from computational-fluid-dynamics approach. As inputs into the data-driven model, the heave and pitch motions of a ship advancing in irregular, i.e., natural seaway were considered. By using different models in the framework of machine learning, required computation for the associated ship motions may be avoided, thus reducing the computational effort to forecast ship motions. Comparative predictions with numerical simulations revealed that the deep-neural-network method for training in auto-machine-learning instructions yielded the highest accuracy in heave motion, resulting in a non-normalized mean-absolute-error of 0.74, against the corresponding error of 1.07 from numerical computations, whereas the method trained with the tree-based models (the extreme gradient boosting and the Hist gradient boosting regressor) predicted less accurate motions for the tested ship. The model trained with the random forest regressor exhibited an error of 1.10. Numerical simulation based on a field method proved to be the most suitable choice for pitch motion. Despite the few samples available to train the regressors, results demonstrated that the measured data was sufficient to assess the developed data-driven model for ship response determination.

Predicting levels of legal case difficulties using machine learning

Lawyers play a crucial role in the courtroom, assisting clients in their defense. Because of their lack of legal expertise, a person or organization facing legal issues requires professional aid. However, we need to know how much money will be spent on paying lawyers. The level of complexity in a case can be used to determine lawyer costs. Therefore, in this research, we propose employing machine learning methodologies, i.e., random forest classifiers and support vector machines (SVM), to determine the level of legal case difficulties. The novelty of this research is applying a machine learning approach in predicting the level of difficulty of legal cases. The data utilized consists of 990 records, which are divided into training and testing data in a 90:10 ratio. The term frequency-inverse document frequency (TF-IDF) approach was then utilized to perform text preprocessing. The text-preprocessing findings are utilized as input in the classification process. According to the research findings, an accuracy value of 85%, a value of weighted average precision is 88%, and a value of weighted average recall is 85%, for support vector machine. Using random forest, we achieve an accuracy value of 89%, a value of weighted average precision is 85.6%, and a value of weighted average recall is 80%.

Predicting coastal wave conditions: A simple machine learning approach

Accurate and reliable nearshore wave predictions are highly valuable for a range of marine activities, including coastal engineering and maritime transport. However, in nearshore locations, predicting wave properties is challenging due to complex shallow water processes, requiring local wave models. This article develops an alternative data-driven framework to predict wave parameters (e.g. significant wave height) through the extension of wave buoy datasets using a trained Gaussian process regression (GPR – a supervised machine learning method). We present an easy-to-implement workflow, where the extensive range of input parameters (from ECMWF’s (1) ERA5 reanalysis and (2) IFS forecast global wave model, ≈50km resolution) drives the development of GPR models. At five contrasting locations around the United Kingdom’s coastline, the GPR models produce wave predictions (forecast and hindcast) with low bias scores and strong correlations with observations. When compared to the global (ERA5 reanalysis) and a benchmark shelf-scale (Atlantic-European North West Shelf reanalysis; AENWS, 1.5−3.0km resolution) model, the GPR hindcasts reduced significant wave height (Hs) root-mean-squared error (RMSE) from 0.46 m (ERA5) and 0.21 m (AENWS) to 0.16 m (GPR). For the average zero-crossing wave period (Tz) RMSE reduced from 1.46 s (ERA5) and 1.15 s (AENWS) to 0.58 s (GPR). Because our approach uses publicly available global data, it can be implemented at any historic or active buoy location. We provide proof of concept for an online forecast and hindcast tool which has the potential to improve accessibility to coastal wave predictions for many marine stakeholders.

Advancing in creep index of soil prediction: A groundbreaking machine learning approach with Multivariate Adaptive Regression Splines

Advancing in creep index of soil prediction: A groundbreaking machine learning approach with Multivariate Adaptive Regression Splines