Supervised Machine Learning Techniques Research Articles

Supervised machine learning for predicting and interpreting dynamic drivers of plantation forest productivity in northern Tasmania, Australia

The capacity of a plantation forest to grow and produce timber is locally constrained by topography, climate, soil conditions, and external factors such as fire and timber harvesting. Accurate estimation of forest productivity supports effective forest management. However, efficiently generating accurate models of forest productivity is hampered by the need to gather, process and integrate large volumes of disparate, high dimensional data that require computationally intensive data analysis processing methods. Recent developments in cloud-based machine learning systems offer a means to address this problem.This research investigates the use of supervised machine learning to model and predict forest productivity across pine (Pinus radiata) plantations in northern Tasmania, Australia. Forest productivity models are generated by integrating 23 predictive features, including multi-temporal LiDAR (Light Detection and Ranging) derived topographic attributes, climate (rainfall and temperature) information, and edaphic conditions (geology and soil). Five machine learning (ML) regression algorithms are compared for this task: Linear Regression (LR), Polynomial Regression (PR), Decision Trees Regression (DT), Random Forests (RF) and Gradient Boosted Decision Trees Regression (GBDT). The best performing algorithm, in terms of the optimal bias-variance trade-off, was RF (RMSE 2.08 and Bias −0.72) followed closely by GBDT (RMSE 2.13 and Bias −0.68) and DT (RMSE 2.94 and Bias −0.68). Tuning Model Complexity was used to provide a clear understanding of the relationship and interactions between input features and forest productivity, resulting in more accurate and interpretable models. In contrast, we conclude that GBDT results in a more reliable predictive performance than RF, and transferability model to unseen data by assessing the spatial autocorrelation. Across the top performing models, rainfall was the most important factor driving forest productivity, followed by geological class, topographic position index (TPI), landscape aspect and Digital Elevation Model (DEM).This work demonstrates the usefulness of supervised machine learning techniques to generate accurate and efficient forest productivity predictions from large and diverse spatial datasets. Moreover, users are afforded the ability to gain insight into changes that affect forest productivity through time, such as the increasing risks of wildfire and climate change by identifying the factors that contribute the most to the tree growth. By delivering a framework to understand complex and dynamic drivers of productivity through a machine learning pipeline enhanced by cloud-based machine learning systems, forest managers are provided with easily accessible tools to maximisation productivity.

Detection of Textual Propaganda Using Passive Aggressive Classifiers

Nowadays, social media activity, particularly news that spreads over the network, is a major source of knowledge. People search out and chew up news from internet-based living because of the low effort, easy access, and rapid dissemination of information. Twitter, as one of the most wellknown continuing news sources, also happens to be one of the most dominant news disseminating media. It has already been known to wreak significant harm by disseminating snippets of gossip. Online clients are typically susceptible, and everything they do on web-based networking media is assumed to be trustworthy. As a result, automating counterfeit propaganda detection is critical to maintaining a vibrant online media and informal organization. In order to computerize propaganda news identification in Twitter datasets, this research develops a technique for recognizing propaganda text messages from tweets by figuring out how to anticipate precision evaluations. This paper proposes a supervised machine learning technique, Passive aggressive classifiers that uses Count Vectorizer and Term FrequencyInverse Document Frequency Vectorizer as feature extraction to detect propaganda news based on the polarity of the corresponding article. Finally, this algorithm uses dataset with 43000 records and shows good accuracy.

Supervised learning algorithms as a tool for archaeology: Classification of ceramic samples described by chemical element concentrations

Ceramic provenance studies often use minor and trace elements to gather knowledge about the presence of local furnaces and commercial trades. There are various chemical techniques that can be used to determine the elemental composition of ceramics, either non-destructively or by requiring samples. From these data, researchers can often determine provenance, and then use multivariate analyses with geological and archaeological information to classify the ceramics. In this study, we aimed to demonstrate the potential of supervised Machine Learning techniques to classify ceramic samples based on their chemical element concentrations. We applied several supervised learning algorithms to a set of 36 fragments whose archaeological classification was already known, using chemical analysis data that had been verified through previous studies. We carried out different sets of experiments, exploiting in different ways the available data, and evaluated the performance of the adopted algorithms, to propose new tools for ceramics provenance studies in archaeology. Our results show that machine learning can be a reliable and useful tool for archaeological classification based on chemical analysis data, providing a reliable and schematic picture of archaeological findings.

Introducing Temporal Correlation in Rainfall and Wind Prediction From Underwater Noise

While in the past the prediction of wind and rainfall from underwater noise was performed using empirical equations fed with very few spectral bins and fitted to the data, it has recently been shown that regression performed using supervised machine learning techniques can benefit from the simultaneous use of all spectral bins, at the cost of increased complexity. However, both empirical equations and machine learning regressors perform the prediction using only the acoustic information collected at the time when one wants to know the wind speed or the rainfall intensity. At most, averages are made between spectra measured at subsequent times (spectral compounding) or between predictions obtained at subsequent times (prediction compounding). In this article, it is proposed to exploit the temporal correlation inherent in the phenomena being predicted, as has already been done in methods that forecast wind and rainfall from their values (and sometimes those of other meteorological quantities) in the recent past. A special architecture of recurrent neural networks, the long short-term memory, is used along with a data set composed of about 16 months of underwater noise measurements (acquired every 10 min, simultaneously with wind and rain measurements above the sea surface) to demonstrate that the introduction of temporal correlation brings significant advantages, improving the accuracy and reducing the problems met in the widely adopted memoryless prediction performed by random forest regression. Working with samples acquired at 10-min intervals, the best performance is obtained by including three noise spectra for wind prediction and six spectra for rainfall prediction.

Machine learning algorithm improves the detection of NASH (NAS-based) and at-risk NASH: A development and validation study.

Detecting NASH remains challenging, while at-risk NASH (steatohepatitis and F≥ 2) tends to progress and is of interest for drug development and clinical application. We developed prediction models by supervised machine learning techniques, with clinical data and biomarkers to stage and grade patients with NAFLD. Learning data were collected in the Liver Investigation: Testing Marker Utility in Steatohepatitis metacohort (966 biopsy-proven NAFLD adults), staged and graded according to NASH CRN. Conditions of interest were the clinical trial definition of NASH (NAS ≥ 4;53%), at-risk NASH (NASH with F ≥ 2;35%), significant (F ≥ 2;47%), and advanced fibrosis (F ≥ 3;28%). Thirty-five predictors were included. Missing data were handled by multiple imputations. Data were randomly split into training/validation (75/25) sets. A gradient boosting machine was applied to develop 2 models for each condition: clinical versus extended (clinical and biomarkers). Two variants of the NASH and at-risk NASH models were constructed: direct and composite models.Clinical gradient boosting machine models for steatosis/inflammation/ballooning had AUCs of 0.94/0.79/0.72. There were no improvements when biomarkers were included. The direct NASH model produced AUCs (clinical/extended) of 0.61/0.65. The composite NASH model performed significantly better (0.71) for both variants. The composite at-risk NASH model had an AUC of 0.83 (clinical and extended), an improvement over the direct model. Significant fibrosis models had AUCs (clinical/extended) of 0.76/0.78. The extended advanced fibrosis model (0.86) performed significantly better than the clinical version (0.82). Detection of NASH and at-risk NASH can be improved by constructing independent machine learning models for each component, using only clinical predictors. Adding biomarkers only improved the accuracy of fibrosis.

Forecasting Student Clothes Purchases Intention in Bangladesh: A Machine Learning Approach

Online shopping provides an excellent opportunity and platform for today's traditional businesses. Because of the advancement of online purchasing systems, students often prefer online shopping. Thus, students' involvement in online purchasing has become an important trend. The research aims to determine university students' purchase intentions toward Bangladeshi clothing brands using several machine learning approaches. An online questionnaire survey was conducted with 1000 university students, and the study goal is to understand their attitudes to online shopping from a different perspective. This paper represents a comparative study of different machine-learning techniques that have been applied to the problem of customer purchasing intention. The experiments were conducted using supervised machine learning techniques like linear regression, logistic regression, and Support Vector Machine (SVM) was also used to predict university students' purchase intentions. This study found that students' age, quality of cloth, purchase discount, and price positively impacted student purchase intentions, but the buying risk negatively affected students' purchase intentions. Linear regression gives the highest accuracy with maximum features, and the accuracy is 89.2%.

Intelligent Sales Prediction Using SVM Algorithm

The Analysis report proposes a solution that grants businesses complete visibility into their inventory and streamlines sales operations from the distribution center to the store. It is used to determine the balance of existing materials when it is demanded to know the stock balance, control the levels of materials, and daily movement of the materials. Data is gathered from the web application and processed to form a group of charts called a dashboard. This dashboard helps to analyze the sales and performance of every product report. Analysis report deals with the organization’s requirements, performance analysis, sales, and financial problems, and solving them, in addition, to saving all the materials information, and process in a private system database. The quality of wood used in the furniture and performance analysis of furniture based on design are represented in form of charts for a better understanding of the user. Admin can view their financial and sales reports by daily, weekly, monthly, and yearly trends. This Furniture Web Application System is created for the convenience of the customers, of this specific shop. It is not only for users but also convenient for the owner to sell their products online along with analytics reports using machine learning techniques. To analyze reports, the system uses Support Vector Machine (SVM), a supervised machine learning technique. Keywords—furniture, machine learning, analysis report

Prediction of Heart Diseases by Using Supervised Machine Learning Algorithms

Heart Disease is a complex and life-threatening ailment that poses a significant mortality risk around the world, with nearly a third of global deaths attributable to heart-related conditions. The early prediction and detection of heart disease are of utmost importance in the medical field, as they may lead to saving numerous lives. However, the lack of heart expertise in many countries and the high rates of misdiagnosis highlight the need for accurate and efficient prediction methods. Machine learning-based approaches have the potential to address this need, particularly in handling the large amounts of data generated by medical sectors and hospitals. In this study, the performance and accuracy of several supervised machine learning algorithms were compared for heart disease prediction using a dataset obtained from PhysioNet databases. The classifiers that were applied included Artificial Neural Network (ANN), Gradient Boosting, Decision Tree, Naive Bayes, and Random Forest. Results showed that the ANN algorithm achieved the highest Accuracy of 94.1%, with a sensitivity and specificity of 94.1%. The study thus concluded that supervised machine learning techniques can be utilized with great success to forecast heart disease, displaying exceptional potential for practical application and accuracy

Stakeholders’ impact on the reuse potential of structural elements at the end-of-life of a building: A machine learning approach

The construction industry, and at its core the building sector, is the largest consumer of non-renewable resources, which produces the highest amount of waste and greenhouse gas emissions worldwide. Since most of the embodied energy and CO2 emissions during the construction and demolition phases of a building are related to its structure, measures to extend the service life of these components should be prioritised. This study develops a set of easy-to-understand instructions to facilitate the practitioners in assessing the social sustainability and responsibility of reusing the load-bearing structural components within the building sector. The results derived by developing and then employing advanced machine learning techniques indicate that the most significant social factor is the perception of the regulatory authorities. The second and third ranks among the social reusability factors belong to risks. Since there is a strong correlation between perception and risk, the potential risks associated with reusing structural elements affect the stakeholders’ perception of reuse. The Bayesian network developed in this study unveil the complex and non-linear correlation between variables, which means none of the factors could alone determine the reusability of an element. This paper shows that by using the basics of probability theory and combining them with advanced supervised machine learning techniques, it is possible to develop tools that reliably estimate the social reusability of these elements based on influencing variables. Therefore, the authors propose using the developed approach in this study to promote materials' circularity in different construction industry sub-sectors.

PREDIKSI JARAK TEMPUH KAPAL MOTOR SANGIANG MENGGUNAKAN SUPERVISED MACHINE LEARNING

As a maritime nation with thousands of islands and a vast sea area, sea transportation is the most effective transportation used by the people of Indonesia. A motorboat is one type of maritime transportation that is used to move people or commodities. In this article, we will discuss predicting the daily mileage of one of the motorboats, the Sangiang, which travels from Bitung to Ternate. Three independent variables, Anchor Time (minutes), Speed (knots/hour), and Sailing Time (minutes), are used in supervised machine learning techniques to estimate the daily mileage (mile). Of the various methods evaluated, the multiple regression model was found to be the most accurate at forecasting the Sangiang motorboat’s daily mileage.

Fractional Order Sliding Mode Controller Based on Supervised Machine Learning Techniques for Speed Control of PMSM

Tracking the speed and current in permanent magnet synchronous motors (PMSMs) for industrial applications is challenging due to various external and internal disturbances such as parameter variations, unmodelled dynamics, and external load disturbances. Inaccurate tracking of speed and current results in severe system deterioration and overheating. Therefore, the design of the controller for a PMSM is essential to ensure the system can operate efficiently under conditions of parametric uncertainties and significant variations. The present work proposes a PMSM speed controller using machine learning (ML) techniques for quick response and insensitivity to parameter changes and disturbances. The proposed ML controller is designed by learning fractional-order sliding mode control (FOSMC) controller behavior. The primary purpose of using ML in FOSMC is to avoid the self-tuning of the parameters and ensure the speed reaches the reference value in finite time with faster convergence and better tracking precision. Furthermore, the ML model does not require the mathematical model of the speed controller. In this work, several ML models are empirically evaluated on their estimation accuracy for speed tracking, namely ordinary least squares, passive-aggressive regression, random forest, and support vector machine. Finally, the proposed controller is implemented on a real-time hardware-in-the-loop (HIL) simulation platform from PLECS Inc. Comparative simulation and experimental results are presented and discussed. It is shown from the comparative study that the proposed FOSMC based on ML outperformed the traditional sliding mode control (SMC), which is more commonly used in industry in terms of tracking speed and accuracy.

Analyzing changes in travel patterns due to Covid-19 using Twitter data in India

Advancements in technology have enabled researchers to gather large-scale mobility information cost-effectively. In fact, with millions of active users, location-based social media (LBSM) platforms such as Facebook, Twitter, Instagram, Flickr, etc., have become potential big data sources to measure individual behaviour. Despite such passive data collection techniques primarily not providing individual-level information, the sheer volume of such data facilitates a better understanding of aggregate patterns. Besides, conducting the conventional transportation survey during the initial waves of the COVID-19 pandemic was nearly impossible due to social- distancing and lockdown rules. In such context, the present research showcases a method for extracting the mobility traces and identifying the travel patterns of visitors and residents in Delhi from geo-labelled posts on Twitter.Initially, a heuristic classification strategy has been developed based on a few spatiotemporal assumptions to identify and differentiate visitors and residents based on user coordinates. Also, three supervised machine learning techniques, i.e., support vector machine (SVM), k-nearest neighbours (kNN) and decision tree, were used to classify users based on their historical coordinates. Afterwards, the spatial variation of their destination preferences was studied using K-Means, DBSCAN, and Means-Shift clustering techniques, out of which the K- Means clustering method performed best. Lastly, the travel patterns from tweets during pre and during pandemic (COVID-19) were compared using respective clusters.We observed that the performance of the proposed heuristic classifier is comparable with the supervised machine learning (ML) technique used for classification. Furthermore, the results indicate that the proposed model can successfully identify the cluster coordinates for visitors' spots as well as the locations of residents. During the pandemic situation, the mean distance travelled by users is significantly reduced. The study also shows that the number of long-distance trips has also decreased. Also, during COVID, tweets were done from very few unique tourist spots. This suggests lower tendencies of people to travel for tourism purposes. The proposed methods for classification and clustering in the present study will be crucial to obtain individual travel patterns from LBSM data.

Can we predict orthodontic extraction patterns by using machine learning?

To investigate the utility of machine learning (ML) in accurately predicting orthodontic extraction patterns in a heterogeneous population. The material of this retrospective study consisted of records of 366 patients treated with orthodontic extractions. The dataset was randomly split into training (70%) and test sets (30%) and was stratified according to race/ethnicity and gender. Fifty-five cephalometric and demographic input data were used to train and test multiple ML algorithms. The extraction patterns were labelled according to the previous treatment plan. Random Forest (RF), Logistic Regression (LR), and Support Vector Machine (SVM) algorithms were used to predict the patient's extraction patterns. The highest class accuracy percentages were obtained for the upper and lower 1st premolars (U/L4s) (RF: 81.63%, LR: 63.27%, SVM: 63.27%) and upper 1st premolars only (U4s) extraction patterns (RF: 61.11%, LR: 72.22%, SVM: 72.22%). However, all methods revealed low class accuracy rates (<50%) for the upper 1st and lower 2nd premolars (U4/L5s), upper 2nd and lower 1st premolars (U5/L4s), and upper and lower 2nd premolars (U/L5s) extraction patterns. For the overall accuracy, RF yielded the highest percentage with 54.55%, followed by SVM with 52.73% and LR with 49.09%. All tested supervised ML techniques yielded good accuracy in predicting U/L4s and U4s extraction patterns. However, they predicted poorly for the U4/L5s, U5/L4s, and U/L5s extraction patterns. Molar relationship, mandibular crowding, and overjet were found to be the most predictive indicators for determining extraction patterns.

Socio-ecological drivers of demersal fishing activity in the North Sea: The case of three German fleets

Worldwide, fisheries face the consequences of climate change and compete with expanding human activities at sea, which may trigger unforeseen reactions of fishers. Hence, knowledge on drivers of fishing behavior is crucial for management and needs to be integrated in resource management policies. In this study, we identify factors influencing fishing activity of North Sea demersal fleets. First, we explore drivers of the North Sea demersal fisheries in scientific literature. Subsequently, we study the effects of identified drivers on the spatio-temporal dynamics of German demersal fisheries using boosted regression trees (BRT), a supervised machine learning technique. An exploratory literature review revealed a lack of studies incorporating biophysical, economic and socio-cultural fishing drivers in a single quantitative analysis. Our BRT analysis contributed to filling this research gap and highlighted the importance of biophysical drivers such as temperature, salinity, and bathymetry for fishing behavior. Contrary to findings of previous studies, our empirical analysis identified quotas and market prices to be irrelevant, except for low brown shrimp prices, which counter-intuitively increased fishing effort. Moreover, economic and socio-cultural variables influencing brown shrimp fishing effort differed from the other fleets, especially determined by increased effort on workdays and reduced effort when fuel prices were high. Our findings provide key information for marine spatial planning and supports the integration of fishing fleet behavior into policies.

Evaluating Various Machine Learning Techniques in Credit Risk Area

Implementing machine learning techniques to credit scoring is a popular method, which is widely used by many financial institutions and banks at present. As the fast development of machine learning tools, these technologies could provide people more accurate predictions and help enterprises avoid future risk. A supervised machine learning technique is utilized in this research as the classification approach. In this experiment, several machine learning algorithms will be compared in order to present the performance by evaluating the type of credit risk. The data is about assessing customers of a German banking systems from the UCI Machine Learning Repository, which contains 5000 instances and 21 attributes. The final result of this research shows the comparison of 12 scenarios among different combinations of balancing methods, feature selection methods, and predictive algorithms, which finally presents that the collection of Adaptive Synthetic, Boruta and k-Nearest Neighbor receives the highest accuracy score.

Prediction of groundwater quality index in the Gaza coastal aquifer using supervised machine learning techniques

Abstract This paper investigates the performance of five supervised machine learning algorithms, including support vector machine (SVM), logistic regression (LogR), decision tree (DT), multiple perceptron neural network (MLP-NN), and K-nearest neighbours (KNN) for predicting the water quality index (WQI) and water quality class (WQC) in the coastal aquifer of the Gaza Strip. A total of 2,448 samples of groundwater were collected from the coastal aquifer of the Gaza Strip, and various physical and chemical parameters were measured to calculate the WQI based on weight. The prediction accuracy was evaluated using five error measures. The results showed that MLP-NN outperformed other models in terms of accuracy with an R value of 0.9945–0.9948, compared with 0.9897–0.9880 for SVM, 0.9784–0.9800 for LogR, 0.9464–0.9247 for KNN, and 0.9301–0.9064 for DT. SVM classification showed that 78.32% of the study area fell under poor to unsuitable water categories, while the north part of the region had good to excellent water quality. Total dissolved solids (TDS) was the most important parameter in WQI predictions while and were the least important. MLP-NN and SVM were the most accurate models for the WQI prediction and classification in the Gaza coastal aquifer.

Real-Time Visual Recognition of Ramp Hand Signals for UAS Ground Operations

We describe the design and validation of a vision-based system that allows the dynamic identification of ramp signals performed by airport ground staff. This ramp signals’ recognizer increases the autonomy of unmanned vehicles and prevents errors caused by visual misinterpretations or lack of attention from the pilot of manned vehicles. This system is based on supervised machine learning techniques, developed with our own training dataset and two models. The first model is based on a pre-trained Convolutional Pose Machine followed by a classifier, for which we have evaluated two possibilities: A Random Forest and a Multi-Layer Perceptron based classifier. The second model is based on a single Convolutional Neural Network that classifies the gestures directly imported from real images. When experimentally tested, the first model proved to be more accurate and scalable than the second one. Its strength relies on a better capacity to extract information from the images and transform the domain of pixels into spatial vectors, which increases the robustness of the classification layer. The second model instead is more adequate for gestures’ identification in low visibility environments, such as during night operations, conditions in which the first model appeared to be more limited, segmenting the shape of the operator. Our results support the use of supervised learning and computer vision techniques for the correct identification and classification of ramp hand signals performed by airport marshallers.

A supervised machine learning technique for combustion diagnosis using a vibration sensor signal

Machine learning (ML) techniques are increasingly spreading in the automotive area. The advantage of using data-based algorithms is evidenced in the study of complex nonlinear phenomena, such as engine combustion. An important aspect when using these techniques is selecting the appropriate features to feed the algorithms. Considering that an accurate estimation of the combustion parameters is necessary to maintain high efficiencies, the following study evaluates the potential of using the engine block vibration signal to estimate parameters, such as the indicated mean effective pressure (IMEP) or the combustion phasing and duration. The block vibration data has proven to contain information regarding engine combustion. Yet, adequate data processing is required to retrieve the features that best contribute to the ML model prediction. To this end, the methodology employs Singular Value Decomposition (SVD) to extract these features from the knock signal’s spectrogram. Then, a correlation with the combustion parameters is made through an artificial neural network (ANN). Results yielded an improvement in the estimation accuracy of the combustion phasing parameters, compared with an ANN model using conventional engine control inputs (spark advance, fuel mass and engine speed). The mean absolute error decreased from 8% for the CA90 up to 25% for the CA50. Furthermore, this approach achieved better generalisation capabilities when unlearned conditions were added to a test set, showing an 80% decrease in the MAE with the proposed method. Measurements were performed on a spark ignition engine at different operating conditions.

Detection of various attacks using B-GNB in health care IoT

Despite various advances in health care applications, Intrusion detection remains a challenging task in wireless network due to numerous amount of traffic and large amount of sensitive data leading to various cyber attacks (DDoS, MITM, scan etc).The sensors are used to collect the real time data of patients or wearable devices and send the same to doctors for analyzing various diseases in early stages using IoT. Wireless network is vulnerable to intruders ,mainly for carrying sensitive information hence this paper proposes supervised machine learning technique B-GNB to detect various attacks in Health care data in IuT.The proposed model chooses the best features and accurately classifies attacks with 96% accuracy.

A critical review of machine learning approaches for predicting diabetes mellitus

Diabetes is a complex, chronic metabolic condition that progresses gradually and causes the patient to have increased blood glucose levels. The three most prevalent kinds of diabetes are type 1, type 2, and gestational. It results in serious health problems such renal problems, heart attacks, foot ulcers, and visual impairment. One of the primary symptoms is hyperglycemia. Early illness progression detection aids in the reduction of additional consequences. The classification of quantitative data most typically uses machine learning techniques. It makes it easier to pinpoint the important elements that are most responsible for the development of diseases. During the pre-processing stage, the least important properties are removed. The classification or forecasting of the condition Diabetes Mellitus using conventional. The discussion and analysis of supervised machine learning techniques focused on their predictive accuracy. The most popular and effective machine learning model among the several extant algorithms is examined using a variety of tools, datasets, and measures. One of the most effective machine learning models is the Deep learning algorithm of artificial neural networks, which is favored by many scientists and researchers in the field of health.