Network-based Machine Learning Model Research Articles

Stock Prediction Using Machine Learning

The Trend of stock price prediction is becoming more popular than ever. Share market is difficult to predict due to its volatile nature. There are no rules to follow to predict what will happen with the stock in the future. To predict accurately is a huge challenge since the market trend always keep changing depending on many factors. The objective is to apply machine learning techniques to predict stocks and maximize the profit. In this work, we have shown that with the help of artificial intelligence and machine learning, the process of prediction can be improved. While doing the literature review, we realized that the most effective machine learning tool for this research include: Artificial Neural Network (ANN), Support Vector Machine (SVM), and Genetic Algorithms (GA). All categories have common and unique findings and limitations. We collected data for about 10 years and used Long Short-Term Memory (LSTM) Neural Network-based machine learning models to analyze and predict the stock price. The Recurrent Neural Network (RNN) is useful to preserve the time-series features for improving profits. The financial data High and Close are used as input for the model.

Machine learning regression approach to on-chip optical frequency combs analyses

We present a practical machine learning (ML) method for serving accessible nonlinear functions, which tackles a regression problem with tremendous parameters. By solving the modified Lugiato–Lefever equation, datasets for emulating the silicon-on-insulator platform and generating the on-chip optical frequency comb (OFC) are gathered. Furthermore, a feed-forward network-based ML model is used to train the datasets, and the prediction of the related parameters is implemented synchronously. Numerical results show that the model combining the finite element method with the ML technique is capable of predicting the properties of on-chip frequency combs for the first time, as far as we know, paving the way for analyzing OFCs based on integrated silicon photonics.

Data-driven application of MEMS-based accelerometers for leak detection in water distribution networks

Water scarcity is a global concern; 68 countries are facing extremely-high to medium-high risk of water stress. In this era of crisis, where water conservation is an absolute necessity, the water distribution networks (WDNs) globally are experiencing significant leaks. These leaks cause tremendous financial loss and unacceptable environmental hazards, thus further aggravating the water scarcity situation. To minimize such damage, the adoption of advanced technologies and methodologies for leak detection in the WDNs is absolutely necessary. In this regard, we have investigated the application of cost-effective MEMS-based accelerometers. Experiments were conducted on real networks (metal and non-metal pipes), over the course of ten months, and the acquired acceleration signals were analyzed using a monitoring algorithm. Monitoring index efficiencies and standard deviations for every leak and no-leak case was extracted. Two individual [KNN and Decision Tree] and two ensembles [Random Forest and Adaboost (Decision Tree)] based machine learning models were developed for the accurate classification of the leak and no-leak cases using extracted features; and separate models were developed for metal and non-metal pipes. Random Forest outperformed the other machine learning models and the overall accuracy reached 100% for metal pipes and 94.93% for non-metal pipes. The machine learning models were further validated using unseen/unlabeled cases and were highly effective in detecting leaks. This study demonstrated the applicability of MEMS-based accelerometers for leak detection and established real network-based machine learning models thereby contributing to the research scarcity in this important area.

Automated predictive analytics tool\xa0for rainfall forecasting

Australia faces a dryness disaster whose impact may be mitigated by rainfall prediction. Being an incredibly challenging task, yet accurate prediction of rainfall plays an enormous role in policy making, decision making and organizing sustainable water resource systems. The ability to accurately predict rainfall patterns empowers civilizations. Though short-term rainfall predictions are provided by meteorological systems, long-term prediction of rainfall is challenging and has a lot of factors that lead to uncertainty. Historically, various researchers have experimented with several machine learning techniques in rainfall prediction with given weather conditions. However, in places like Australia where the climate is variable, finding the best method to model the complex rainfall process is a major challenge. The aim of this paper is to: (a) predict rainfall using machine learning algorithms and comparing the performance of different models. (b) Develop an optimized neural network and develop a prediction model using the neural network (c) to do a comparative study of new and existing prediction techniques using Australian rainfall data. In this paper, rainfall data collected over a span of ten years from 2007 to 2017, with the input from 26 geographically diverse locations have been used to develop the predictive models. The data was divided into training and testing sets for validation purposes. The results show that both traditional and neural network-based machine learning models can predict rainfall with more precision.

Machine learning-based data processing technique for time-domain thermoreflectance (TDTR) measurements

Machine learning (ML) has emerged as an increasingly important research tool and has shown great potential for efficient and high-throughput experimental data processing. Meanwhile, ultrafast laser-based time-domain thermoreflectance (TDTR) has been developed into a powerful thermal characterization technique and has been widely applied to measure thermal properties of both bulk and thin-film materials. In this work, artificial neural network-based ML models have been trained for data processing in TDTR experiments. One generally applicable ML model could be trained to process the experimental data of different samples measured using different modulation frequencies and laser spot sizes. Our results suggest that ML is not only fast and efficient in data processing but also accurate and powerful, capable of detecting minute features in the experimental signals and thus enabling extraction of multiple (three or more) parameters simultaneously from the experimental data. The ML model also enables high-speed estimation of the uncertainties of multiple parameters using the Monte Carlo method.

Application of New Artificial Neural Network to Predict Heat Transfer and Thermal Performance of a Solar Air-Heater Tube

In the present study, the heat transfer and thermal performance of a helical corrugation with perforated circular disc solar air-heater tubes are predicted using a machine learning regression technique. This paper describes a statistical analysis of heat transfer by developing an artificial neural network-based machine learning model. The effects of variation in the corrugation angle (θ), perforation ratio (k), corrugation pitch ratio (y), perforated disc pitch ratio (s), and Reynolds number have been analyzed. An artificial neural network model is used for regression analysis to predict the heat transfer in terms of Nusselt number and thermohydraulic efficiency, and the results showed high prediction accuracies. The artificial neural network model is robust and precise, and can be used by thermal system design engineers for predicting output variables. Two different models are trained based on the features of experimental data, which provide an estimation of experimental output based on user-defined input parameters. The models are evaluated to have an accuracy of 97.00% on unknown test data. These models will help the researchers working in heat transfer enhancement-based experiments to understand and predict the output. As a result, the time and cost of the experiments will reduce.

A patient network-based machine learning model for disease prediction: The case of type 2 diabetes mellitus

A patient network-based machine learning model for disease prediction: The case of type 2 diabetes mellitus

Machine learning models to quantify HER2 for real-time tissue image analysis in prospective clinical trials.

3061 Background: Patient eligibility for HER2-targeting treatments is commonly informed by testing tumor HER2 expression using immunohistochemistry. As HER2 expression is visually assessed by pathologists, inter- and intra-rater variability might affect treatment decisions. Here, we report the development of an automated machine learning (ML)-based algorithm to quantify HER2 cell membrane expression across a diversity of breast cancer phenotypes as a clinical tool for monitoring HER2 testing quality. Methods: A total of 689 breast cancer tissue samples were either procured (Avaden Biosciences) or were anonymized samples from the AstraZeneca biobank comprising tissues from primary and metastatic tumors, core needle biopsies and surgical resections, lobular and ductal carcinomas, across tumor grades and HER2 expression levels. Samples were stained for HER2 detection (Ventana HER2 (4B5) Assay) and digitized (Leica Biosystems) across 5 laboratories in the US. Whole-slide images (WSIs) were stratified into training (n = 407), validation (n = 110), and test sets (n = 172). Multiple convolutional neural network based ML models (PathAI, Boston, MA) were trained using 190,000 manual annotations provided by 30 board-certified pathologists to identify artifacts, invasive tumor, identify individual cancer cells and measure tumor cell membrane HER2 expression as partial or complete, and negative, weak-or-moderate, or intense. Cell-level scores were validated against a consensus of manual cell counts from 5 independent pathologists in 320 representative regions of test set WSIs. HER2 scores were generated by automatically applying rules derived from 2018 ASCO/CAP guidelines and then compared in the test set with consensus scores from 3 independent pathologists. Results: Cell counts provided by the ML model were strongly consistent with cell counts obtained by pathologist consensus in all cell-types except for faintly positive HER2 cells where ML-based quantification identified more cells on average. Automatically generated ML-ASCO/CAP HER2 scores using WSI showed substantial consistency across IHC categories with the consensus of pathologists (ICC 0.88, 95%CI 0.82-0.92) in the test set and improved further when ML models were trained to agree with pathologists by adjusting cut offs (ICC 0.91, 95%CI 0.89-0.94). The ML-based model was deployed through the PathAI cloud platform to calculate HER2 testing quality control metrics in real-time in multicentric clinical trials. Conclusions: Automated image analysis of HER2-stained breast cancer tissues using ML-based models is consistent with pathologist consensus across breast cancer tissue types. The results support evidence that ML-based algorithms can help pathologists assess HER2 testing reproducibility in clinical trials.

Application of network link prediction in drug discovery

BackgroundTechnological and research advances have produced large volumes of biomedical data. When represented as a network (graph), these data become useful for modeling entities and interactions in biological and similar complex systems. In the field of network biology and network medicine, there is a particular interest in predicting results from drug–drug, drug–disease, and protein–protein interactions to advance the speed of drug discovery. Existing data and modern computational methods allow to identify potentially beneficial and harmful interactions, and therefore, narrow drug trials ahead of actual clinical trials. Such automated data-driven investigation relies on machine learning techniques. However, traditional machine learning approaches require extensive preprocessing of the data that makes them impractical for large datasets. This study presents wide range of machine learning methods for predicting outcomes from biomedical interactions and evaluates the performance of the traditional methods with more recent network-based approaches.ResultsWe applied a wide range of 32 different network-based machine learning models to five commonly available biomedical datasets, and evaluated their performance based on three important evaluations metrics namely AUROC, AUPR, and F1-score. We achieved this by converting link prediction problem as binary classification problem. In order to achieve this we have considered the existing links as positive example and randomly sampled negative examples from non-existant set. After experimental evaluation we found that Prone, ACT and LRW_5 are the top 3 best performers on all five datasets.ConclusionsThis work presents a comparative evaluation of network-based machine learning algorithms for predicting network links, with applications in the prediction of drug-target and drug–drug interactions, and applied well known network-based machine learning methods. Our work is helpful in guiding researchers in the appropriate selection of machine learning methods for pharmaceutical tasks.

Machine learning discovery of distinguishing laboratory features for severity classification of COVID‐19 patients

The exponential spread of COVID‐19 worldwide is evident, with devastating outbreaks primarily in the United States, Spain, Italy, the United Kingdom, France, Germany, Turkey and Russia. As of 1 May 2020, a total of 3,308,386 confirmed cases have been reported worldwide, with an accumulative mortality of 233,093. Due to the complexity and uncertainty of the pathology of COVID‐19, it is not easy for front‐line doctors to categorise severity levels of clinical COVID‐19 that are general and severe/critical cases, with consistency. The more than 300 laboratory features, coupled with underlying disease, all combine to complicate proper and rapid patient diagnosis. However, such screening is necessary for early triage, diagnosis, assignment of appropriate level of care facility, and institution of timely intervention. A machine learning analysis was carried out with confirmed COVID‐19 patient data from 10 January to 18 February 2020, who were admitted to Tongji Hospital, in Wuhan, China. A softmax neural network‐based machine learning model was established to categorise patient severity levels. According to the analysis of 2662 cases using clinical and laboratory data, the present model can be used to reveal the top 30 of more than 300 laboratory features, yielding 86.30% blind test accuracy, 0.8195 F1‐score, and 100% consistency using a two‐way patient classification of severe/critical to general. For severe/critical cases, F1‐score is 0.9081 (i.e. recall is 0.9050, and precision is 0.9113). This model for classification can be accomplished at a mini‐second‐level computational cost (in contrast to minute‐level manual). Based on available COVID‐19 patient diagnosis and therapy, an artificial intelligence model paradigm can help doctors quickly classify patients with a high degree of accuracy and 100% consistency to significantly improve diagnostic and classification efficiency. The discovered top 30 laboratory features can be used for greater differentiation to serve as an essential supplement to current guidelines, thus creating a more comprehensive assessment of COVID‐19 cases during the early stages of infection. Such early differentiation will help the assignment of the appropriate level of care for individual patients.

WITHDRAWN: A neural network based machine learning model in digital health care for wait-time prediction

It is an inevitable part of life to wait in queues, and not knowing how long the wait would last can be a major concern. In order to minimise this, businesses also try to forecast waiting times in order to be able to handle their sleeves. In medical care, this is extremely important because the patients are possibly already in some pain. This study compares the performance of two separate ML methods and a simulation approach to the issue of wait-time prediction in digital healthcare environments. In order to match the best ML model with the simulation approach, a hybrid method was also introduced. The ML approaches used historical data of the patient queue to build a model for new patients who enter the queue to estimate the waiting time. The simulation algorithm imitates the queue in a simulated world and simulates time going forward to allocate a clinician to the new patient who enters the queue and thus creates a wait time assessment. As an additional feature for the best ML model, the simulation approach used the wait-time estimates provided by the simulation algorithm. The Machine Learning sequence modelling approach was implemented and defined by a Time Convolutional Network (TCN) model and a Long Short Term Memory Network (LSTM) model. The conventional approach to ML was implemented as a Random Forest Regressor (RF) model and a Support Vector Regressor (SVR) model. The exponential smoothing pre-processing technique was used to incorporate the temporal dimension into the conventional ML approach. The findings showed that all models vary statistically significantly. The TCN model and simulation algorithm had all individual models with the lowest mean square error (MSE). Compared to both conventional ML models, both sequence models had lower MSE. The MSE model had the lowest MSE of the whole and had both the ML and the simulation method the highest output characteristics. The hybrid model is however the most complex and thus needs the most maintenance.

Machine learned features from density of states for accurate adsorption energy prediction

Materials databases generated by high-throughput computational screening, typically using density functional theory (DFT), have become valuable resources for discovering new heterogeneous catalysts, though the computational cost associated with generating them presents a crucial roadblock. Hence there is a significant demand for developing descriptors or features, in lieu of DFT, to accurately predict catalytic properties, such as adsorption energies. Here, we demonstrate an approach to predict energies using a convolutional neural network-based machine learning model to automatically obtain key features from the electronic density of states (DOS). The model, DOSnet, is evaluated for a diverse set of adsorbates and surfaces, yielding a mean absolute error on the order of 0.1 eV. In addition, DOSnet can provide physically meaningful predictions and insights by predicting responses to external perturbations to the electronic structure without additional DFT calculations, paving the way for the accelerated discovery of materials and catalysts by exploration of the electronic space.

Backpropagation Neural Network-Based Machine Learning Model for Prediction of Blood Urea and Glucose in CKD Patients.

Diabetes mellitus and its complication such as heart disease, stroke, kidney failure, etc. is a serious concern all over the world. Hence, monitoring some important blood parameters non-invasively is of utmost importance, that too with high accuracy. This paper presents an in-house developed system, which will be helpful for diabetes patients with Chronic Kidney Disease (CKD) to monitor blood urea and glucose. This manuscript discusses a comparative study for the prediction of blood urea and glucose using Backpropagation Artificial Neural Network (BP- ANN) and Partial Least Square Regression (PLSR) model. The NVIDIA Jetson Nano board controls the five fixed LED wavelengths in the Near Infrared (NIR) region from n}{}2.0~mu text{m} to n}{}2.5~mu text{m}n with a constant emission power of 1.2 mW. The spectra for 57 laboratory prepared samples conforming with major blood constituents of the blood sample were recorded. From these samples, 53 spectra were used for training/calibration of the BP-ANN/PLSR model and the remaining 4 samples were used for validating the model. The PLSR model predicts blood urea and glucose with a Root Mean Square Error (RMSE) of 0.88 & 12.01 mg/dL, Coefficient of Determination R2 = 0.93 & R2 = 0.97, Accuracy of 94.2 % and 90.14 %, respectively. To improve the prediction accuracy, BP-ANN model is applied. Later the Principal Component Analysis (PCA) technique was applied to these 57 spectra values. These PCA values were used to train and validate the BP-ANN model. After applying the BP-ANN model, the prediction of blood urea & glucose improved remarkably, which achieved RMSE of 0.69 mg/dL, R2 = 0.96, Accuracy of 95.96 % for urea and RMSE of 2.06 mg/dL, R2 = 0.99, and Accuracy of 98.65 % for glucose. The system performance is then evaluated with Bland Altman analysis and Clarke Error Grid Analysis (CEGA).

Backpropagation Neural Network-Based Machine Learning Model for Prediction of Soil Friction Angle

In the design process of foundations, pavements, retaining walls, and other geotechnical matters, estimation of soil strength-related parameters is crucial. In particular, the friction angle is a critical shear strength factor in assessing the stability and deformation of geotechnical structures. Practically, laboratory or field tests have been conducted to determine the friction angle of soil. However, these jobs are often time-consuming and quite expensive. Therefore, the prediction of geo-mechanical properties of soils using machine learning techniques has been widely applied in recent times. In this study, the Bayesian regularization backpropagation algorithm is built to predict the internal friction angle of the soil based on 145 data collected from experiments. The performance of the model is evaluated by three specific statistical criteria, such as the Pearson correlation coefficient (R), root mean square error (RMSE), and mean absolute error (MAE). The results show that the proposed algorithm performed well for the prediction of the friction angle of soil (R = 0.8885, RMSE = 0.0442, and MAE = 0.0328). Therefore, it can be concluded that the backpropagation neural network-based machine learning model is a reasonably accurate and useful prediction tool for engineers in the predesign phase.

Complex temperature dependence of coherent and incoherent lattice thermal transport in superlattices

Currently, it is still unclear how and to what extent a change in temperature impacts the relative contributions of coherent and incoherent phonons to thermal transport in superlattices. Some seemingly conflicting computational and experimental observations of the temperature dependence of lattice thermal conductivity make the coherent–incoherent thermal transport behaviors in superlattices even more elusive. In this work, we demonstrate that incoherent phonon contribution to thermal transport in superlattices increases as the temperature increases due to elevated inelastic interfacial transmission. On the other hand, the coherent phonon contribution decreases at higher temperatures due to elevated anharmonic scattering. The competition between these two conflicting mechanisms can lead to different trends of lattice thermal conductivity as temperature increases, i.e. increasing, decreasing, or non-monotonic. Finally, we demonstrate that the neural network-based machine learning model can well capture the coherent–incoherent transition of lattice thermal transport in the superlattice, which can greatly aid the understanding and optimization of thermal transport properties of superlattices.

Statistical and artificial neural network-based analysis to understand complexity and heterogeneity in preeclampsia

Preeclampsia is a pregnancy associated disease. It is characterized by high blood pressure and symptoms that are indicative of damage to other organ systems, most often involving the liver and kidneys. If left untreated, the condition could be fatal to mother and baby. This makes it important to delineate the complexities associated with the disease at a molecular level that would help develop methods for early diagnosis. In microarray-based studies, Textoris et al. and Mirzakhani et al. have analyzed the transcriptome with a view to identify biomarkers for preeclampsia. The current study has extensively analyzed these microarray data sets to understand the complexity and heterogeneity associated with preeclampsia. A statistical multiple comparisons-based approach has been used to identify features capable of distinguishing preeclampsia from normotensive cases. These features were then used to build an artificial neural network-based machine learning model that successfully classified the samples. Further, the machine learning model was used to delineate features critical for its internal representation by extending the calliper randomization approach to the analysis of microarray data. Functional analysis of the features identified by the calliper randomization approach revealed pathways that could be crucially involved in the mechanism of the underlying disease. Biological processes associated with the features identified have revealed among others, genes involved in reproductive processes to be differentially expressed.

EVALUATION OF DRUG ADDICTION CAUSES IN PUNJAB USING MACHINE LEARNING

Nowadays, one of the most prevalent issues of every nation in the world is drug addiction. The issue of drug abuse is more common in youth of age category from 15 to 45. Punjab is no exception to this fact. This paper evaluates various causes behind the drug addiction among Punjabi youth and proposed a machine learning model to identify the key parameters for the agony of Punjab. The pre-processing of proposed model involves data extraction from web sources and cleaning of data through basic Natural Language Processing techniques. The post-processing phase contains training and testing of neural network based machine learning model for the classification of Punjabi drug abusers. The proposed method recorded highest accuracy of classification among other benchmarks available in the literature.

Artificial neural network to predict the degraded mechanical properties of metallic materials due to the presence of hydrogen

Machine learning models were introduced to develop a relationship between the elemental composition and degraded mechanical properties in metallic materials due to the presence of hydrogen. Single layer and multilayer feed forward back propagation algorithm was developed as artificial neural network based machine learning models to predict the mechanical properties of hydrogen charged metallic materials. Multilayer feed forward back propagation model was used to predicts the tensile strength, had a network topology of 12-13-3-2. And the single layer feed forward back propagation model was employed to predict the percentage of elongation, has a network topology of 12-11-1. The developed models were validated and tested with unknown inputs and their capability was studied. The models were evaluated using Mean Absolute (MAE) value and represented the scatter diagram to demonstrate the efficiency of the models. The R-value for both the models seems to prove that the models are ready to be used in the practical applications.

Short-term speed predictions exploiting big data on large urban road networks

Big data from floating cars supply a frequent, ubiquitous sampling of traffic conditions on the road network and provide great opportunities for enhanced short-term traffic predictions based on real-time information on the whole network. Two network-based machine learning models, a Bayesian network and a neural network, are formulated with a double star framework that reflects time and space correlation among traffic variables and because of its modular structure is suitable for an automatic implementation on large road networks. Among different mono-dimensional time-series models, a seasonal autoregressive moving average model (SARMA) is selected for comparison. The time-series model is also used in a hybrid modeling framework to provide the Bayesian network with an a priori estimation of the predicted speed, which is then corrected exploiting the information collected on other links. A large floating car data set on a sub-area of the road network of Rome is used for validation. To account for the variable accuracy of the speed estimated from floating car data, a new error indicator is introduced that relates accuracy of prediction to accuracy of measure. Validation results highlighted that the spatial architecture of the Bayesian network is advantageous in standard conditions, where a priori knowledge is more significant, while mono-dimensional time series revealed to be more valuable in the few cases of non-recurrent congestion conditions observed in the data set. The results obtained suggested introducing a supervisor framework that selects the most suitable prediction depending on the detected traffic regimes.