Azure ML Research Articles

클라우드 기반 플랫폼의 분석 자동화 기술 동향

The demand for data analytics continues to grow as the number of citizen data analysts engaging in analytics activities increases. Data analysis requires a number of complex procedures, all of which can be resource-intensive for analysts to perform manually. This paper compares the technology trends of major cloud-based solutions and discusses automation strategies and technology trends in the cloud environment, with a focus on balancing analytical autonomy and automation, portability of the analytics environment, and improving platform usability. In particular, it compares and analyzes representative platforms such as SageMaker, Vertex AI, and Azure ML to understand the evolution of analytics automation technology and discusses essential components for future advancements. It also discusses how organizations can increase productivity through analytics by using automated platforms for data-driven decision-making.

Monitoring and Prediction of Indoor Air Quality for Enhanced Occupational Health

The amount of moisture in the air is represented by relative humidity (RH); an ideal level of humidity in the interior environment is between 40% and 60% at temperatures between 18° and 20° Celsius. When the RH falls below this level, the environment becomes dry, which can cause skin dryness, irritation, and discomfort at low temperatures. When the humidity level rises above 60%, a wet atmosphere develops, which encourages the growth of mold and mites. Asthma and allergy symptoms may occur as a result. Human health is harmed by excessive humidity or a lack thereof. Dehumidifiers can be used to provide an optimal level of humidity and a stable and pleasant atmosphere; certain models disinfect and purify the water, reducing the spread of bacteria. The design and implementation of a client-server indoor and outdoor air quality monitoring application are presented in this paper. The Netatmo station was used to acquire the data needed in the application. The client is an Android application that allows the user to monitor air quality over a period of their choosing. For a good monitoring process, the Netatmo modules were used to collect data from both environments (indoor: temperature (T), RH, carbon dioxide (CO2), atmospheric pressure (Pa), noise and outdoor: T and RH). The data is stored in a database, using MySQL. The Android application allows the user to view the evolution of the measured parameters in the form of graphs. Also, the paper presents a prediction model of RH using Azure Machine Learning Studio (Azure ML Studio). The model is evaluated using metrics: Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Relative Absolute Error (RAE), Relative Squared Error (RSE) and Coefficient of Determination (CoD).

Process Parameter Optimization of Additively Manufactured Parts Using Intelligent Manufacturing

Additive manufacturing is the technique of combining materials layer by layer and process parameter optimization is a method used popularly for achieving the desired quality of a part. In this paper, four input parameters (layer height, infill density, infill pattern, and number of perimeter walls) along with their settings were chosen to maximize the tensile strength for a given part. Taguchi DOE was used to generate an L27 orthogonal array which helped to fabricate 27 parts on the Ender 3 V2 fused deposition modeling (FDM) printer. The ultimate testing machine was used to test all 27 samples to generate the respective tensile strength values. Next, the Microsoft Azure ML database was used to predict the values of the tensile strength for various input parameters by using the data obtained from Taguchi DOE as the input. Linear regression was applied to the dataset and a web service was deployed through which an API key was generated to find the optimal values for both the input and output parameters. The optimum value of tensile strength was 22.69 MPa at a layer height of 0.28 mm, infill density of 100%, infill pattern of honeycomb, and the number of perimeter walls as 4. The paper ends with the conclusions drawn and future research directions.

Consideration of the possibilities of applying machine learning methods for data analysis when promoting services to bank's clients

The object of the research is modern online services and machine learning libraries for predicting the probability of the bank client's consent to the provision of the proposed services. One of the most problematic areas is the high unpredictability of the result in the field of banking marketing using the most common technique of introducing new services for clients – the so-called cold calling. Therefore, the question of assessing the probability and predicting the behavior of a potential client when promoting new banking services and services using cold calling is particularly relevant. In the course of the study, libraries of machine learning methods and data analysis of the Python programming language were used. A program was developed to build a model for predicting the behavior of bank customers using data processing methods using gradient boosting, regularization of gradient boosting, random forest algorithm and recurrent neural networks. Analogous models were built using cloud machine learning services Azure ML, BigML and the Auto-sklearn library. Data analysis and prediction models built using Python language libraries have a fairly high quality – an average of 94.5 %. Using the Azure ML cloud service, a predictive model with an accuracy of 88.6 % was built. The BigML machine learning service made it possible to build a model with an accuracy of 88.8 %. Machine learning methods from the Auto-sklearn library made it possible to obtain a model with a higher quality – 94.9 %. This is due to the fact that the proposed libraries of the Python programming language allow better customization of data processing methods and machine learning to obtain more accurate models than free cloud services that do not provide such capabilities. Thanks to this, it is possible to obtain a predictive model of the behavior of bank customers with a fairly high degree of accuracy. It is worth noting that in order to make a prediction (forecast), it is necessary to study the context of the task, process the data, build various machine learning algorithms, evaluate the quality of the models and choose the best of them.

On the Performance of Machine Learning Models for Anomaly-Based Intelligent Intrusion Detection Systems for the Internet of Things

Anomaly-based machine learning-enabled intrusion detection systems (AML-IDSs) show low performance and prediction accuracy while detecting intrusions in the Internet of Things (IoT) than that of deep learning-based intrusion detection systems (DL-IDSs). In particular, AML-IDS that employ low complexity models for IoT, such as the principal component machine (PCA) method and the one-class support vector machine (1-SVM) method, are inefficient in detecting intrusions when compared to DL-IDS with the two-class neural network (2-NN) method. PCA and 1-SVM AML-IDS suffer from low detection rates compared to DL-IDS. The size of the data set and the number of features or variants in the data set may influence how well PCA and 1-SVM AML-IDS perform compared to DL-IDS. We attribute the low performance and prediction accuracy of the AML-IDS model to an imbalanced data set, a low similarity index between the training data and testing data, and the use of a single-learner model. The intrinsic limitations of the single-learner model have a direct impact on the accuracy of an intelligent IDS. Also, the dissimilarity between testing data and training data leads to an increasingly high rate of false positives (FPs) in AML-IDS than DL-IDS, which have low false alarms and high predictability. In this article, we examine the use of optimization techniques to enhance the performance of single-learner AML-IDS, such as PCA and 1-SVM AML-IDS models for building efficient, scalable, and distributed intelligent IDS for detecting intrusions in IoT. We evaluate these AML-IDS models by tuning hyperparameters and ensemble learning optimization techniques using the Microsoft Azure ML Studio (AMLS) platform and two data sets containing malicious and benign IoT and industrial IoT (IIoT) network traffic. Furthermore, we present a comparative analysis of AML-IDS models for IoT regarding their performance and predictability.

Using machine learning algorithms in Microsoft Azure ML to improve system search

Using machine learning algorithms in Microsoft Azure ML to improve system search

Azure machine learning studio and SCADA data for failure detection and prediction purposes: A case of wind turbine generator

Most industrial systems have supervisory control and data acquisition (SCADA) systems that collect and store process parameters. SCADA data is seen as a valuable source to get and extract insights about the asset health condition and associated maintenance operations. It is still unclear how appliable and valid insights SCADA data might provide. The purpose of this paper is to explore the potential benefits of SCADA data for maintenance purposes and discuss the limitations from a machine learning perspective. In this paper, a two-year SCADA data related to a wind turbine generator is extracted and analysed using several machine learning algorithms, i.e., two-class boosted decision tree, two-class decision forest, k-means clustering on Azure ML learning studio. It is concluded that the SCADA data can be useful for failure detection and prediction once rich training data is given. In a failure prediction context, data richness means ensuring that fault features are presented in the training data. Moreover, the logs file can be used as labelled data to supervise some algorithms once they are reported in a more rigorous manner (timing, description).

Activity Recognition and Creation of Web Service for Activity Recognition using Mobile Sensor Data using Azure Machine Learning Studio

With the increasing pervasive computation, “Activity Recognition” has become a vast and popular field of research. In the field of automated Activity Recognition, we use multiple sensors in wearable/portable devices in order to recognize the human activities such as standing still, sitting, relaxing, laying, walking, climbing stairs, knee bending cycling jogging etc. The main purpose of this paper is to discuss the field of Activity Recognition for patients and old- age persons or any person in general. This research paper can also be used for telemedicine purposes. Besides, different machine learning algorithm will be applied to achieve Activity Recognition rather precisely. Microsoft Azure ML Studio and a bench marking data set are used for creation as well as evaluation of Machine Learning Model. In addition, a Web Service for Activity Recognition is also developed by using Microsoft Azure ML Studio in order to help the developer and researcher while working on Activity Recognition.

Critique of Various Algorithms for Handwritten Digit Recognition Using Azure ML Studio

Handwritten Digit Recognition is probably one of the most exciting works in the field of science and technology as it is a hard task for the machines to recognize the digits which are written by different people. The handwritten digits may not be perfect and also consist of different flavors. And there is a necessity for handwritten digit recognition in many real-time purposes. The widely used MNIST dataset consists of almost 60000 handwritten digits. And to classify these kinds of images, many machine learning algorithms are used. This paper presents an in-depth analysis of accuracies and performances of Support Vector Machines (SVM), Neural Networks (NN), Decision Tree (DT) algorithms using Microsoft Azure ML Studio.

A Study on Comparison of Lung Cancer Prediction Using Ensemble Machine Learning

Lung cancer is a chronic disease which ranks fourth in cancer incidence with 11 percent of the total cancer incidence in Korea. To deal with such issues, there is an active study on the usefulness and utilization of the Clinical Decision Support System (CDSS) which utilizes machine learning. Thus, this study reviews existing studies on artificial intelligence technology that can be used in determining the lung cancer, and conducted a study on the applicability of machine learning in determination of the lung cancer by comparison and analysis using Azure ML provided by Microsoft. The results of this study show different predictions yielded by three algorithms: Support Vector Machine (SVM), Two-Class Support Decision Jungle and Multiclass Decision Jungle. This study has its limitations in the size of the Big data used in Machine Learning. Although the data provided by Kaggle is the most suitable one for this study, it is assumed that there is a limit in learning the data fully due to the lack of absolute figures. Therefore, it is claimed that if the agency's cooperation in the subsequent research is used to compare and analyze various kinds of algorithms other than those used in this study, a more accurate screening machine for lung cancer could be created.

Cloud-based Non-intrusive Load Monitoring System (NILM)

Design and development of a cloud-based non-intrusive load monitoring System (NILM) is presented. It serves for monitoring and disaggregating the aggregated data such as smart metering into appliance-level load information by using cloud computing and machine learning algorithms implemented in cloud. The existing NILM systems are lack of scalability and limited in computing resources (computation and data storage) due to dedicated, closed and proprietary-based characteristics. They are inaccessible to variety of heterogeneity data (electrical and non-electrical data) openly for improving NILM performance. Therefore, this paper proposed a novel cloud-based NILM system to enable collection of these open data for load monitoring and other energy-related services. The collected data such as smart meter or data acquisition unit (DAQ), is pre-processed and uploaded to the cloud platform. A classifier algorithm based on Artificial Neural Network (ANN) is implemented in Azure ML Studio (AzureML), followed by the classifier testing with different combinations of feature set for the performance comparison. Furthermore, a web service is deployed for web APIs (Application Programming Interfaces) of applications such as smart grid and smart cities. The results shows that the ANN classifier for multiclass classification has improved performance with additional features of harmonics, apart from active and reactive powers used. It also demonstrates the feasibility of proposed cloud-based classifier model for load monitoring. Therefore, the proposed solution offers a convenient and cost-effective way of load monitoring via cloud computing technology for smart grid and smart home applications. Further work includes the use of other ML algorithms for classifier, performance analysis, development of cloud-based universal appliance data and use cases.

Sales prediction of Four Wheelers Unit (4W) with seasonal algorithm Trend Decomposition with Loess (STL) in PT. Astra International, Tbk

Corporate Information System and Technology (CIS&T) division in PT Astra International Tbk is digitizing all business lines. Therefore, the company started to utilize machine learning to predict four wheelers (4W) sales in 2017. Data used were car sales data from 2014 to 2016. The data were obtained from company’s data warehouse (DW). Further, to develop application, firstly, Minitab was used to compare algorithm in modelling phase. Then, Azure Machine Learning Studio (Azure ML) was utilized to build prediction model based on algorithm chose in first phase. Finally, visualization was developed using Power Business Intelligence (Power BI). Time series algorithm used to solve the case is Seasonal Trend Decomposition with Loess (STL). While, machine learning model uses R library, and process flow module in Azure ML. The STL prediction model yield 9.38% error. Visualization of result is divided into two sections filtered by three parameters type, color and area. The first section in dashboard is sales data report showing percentage of type, color of cars and deployment of 4W units in Indonesia through ArcGIS Maps. The second section is forecast result containing chart of x axis to represent year and y axis to represent number of units. Further, the chart provide historical, current and forecast information. The research contributes to illustrate development of prediction using machine learning in industrial environment.

Performance Comparison of Machine Learning Platforms

In this paper, we present a method for comparing and evaluating different collections of machine learning algorithms on the basis of a given performance measure (e.g., accuracy, area under the curve (AUC), F-score). Such a method can be used to compare standard machine learning platforms such as SAS, IBM SPSS, and Microsoft Azure ML. A recent trend in automation of machine learning is to exercise a collection of machine learning algorithms on a particular problem and then use the best performing algorithm. Thus, the proposed method can also be used to compare and evaluate different collections of algorithms for automation on a certain problem type and find the best collection. In the study reported here, we applied the method to compare six machine learning platforms – R, Python, SAS, IBM SPSS Modeler, Microsoft Azure ML, and Apache Spark ML. We compared the platforms on the basis of predictive performance on classification problems because a significant majority of the problems in machine learning are of that type. The general question that we addressed is the following: Are there platforms that are superior to others on some particular performance measure? For each platform, we used a collection of six classification algorithms from the following six families of algorithms – support vector machines, multilayer perceptrons, random forest (or variant), decision trees/gradient boosted trees, Naive Bayes/Bayesian networks, and logistic regression. We compared their performance on the basis of classification accuracy, F-score, and AUC. We used F-score and AUC measures to compare platforms on two-class problems only. For testing the platforms, we used a mix of data sets from (1) the University of California, Irvine (UCI) library, (2) the Kaggle competition library, and (3) high-dimensional gene expression problems. We performed some hyperparameter tuning on algorithms wherever possible. The online supplement is available at https://doi.org/10.1287/ijoc.2018.0825 .

Abstract P2-08-52: A predictive model for distant metastasis in breast cancer patients using machine learning

Abstract Introduction Tumor metastasis is a major clinical challenge accounting for the vast majority of cancer related deaths.In previous studies, prediction of distant metastasis was based on subtypes,clinical status and sometimes gene expression were used however clinical application was difficult. In this study, we develop the easy to use prediction tool for distant metastasis using clinical characteristics and gene profiles which came from CancerSCANTM, Next Generation Sequencing based targeted-sequencing platform designed at Samsung Medical Center(SMC). Methods We performed a retrospective chart review of 326 breast cancer patients who underwent surgery and CancerSCAN TM between Jan 2001 and Dec 2014 at SMC. Median follow up period was 83 months (Range 1˜190). Cancer scanTM cover 381 genes but 27 genes and 34 occasions (loss of function, mutation or copy number variation) were selected for analysis through gradient boosting and Wilcoxon Signed rank test. Azure Machine Learning is a cloud service that enables the execution of machine learning processes.This was accomplished using the steps of (1) edit the data, (2) split the data, (3) train the model, (4) score the model, and (5) evaluate the model. We split the modeling data into training and testing sets using a randomized 50–50 split. Two-class Decision Forest method was used. After deploying the Azure ML predictive model as a web service, we used a Representational State Transfer application programming interface to send data and obtained predictions in real-time. Results No distant metastasis group and distant metastasis group consisted of 267 and 59 patients, respectively. HR-/HER2+ and 50 years old and over patients were higher in metastasis group (p-value = 0.003 and p-value = 0.000). Nuclear grade 3 and N2,3 were higher in metastasis group (p-value = 0.010 and p-value = 0.000, p-value = 0.001 respectively). Stage III was also higher in metastasis group (p-value = 0.000). Among 59 patients with distantmetastasis, multiple sites metastasis was 21 cases (35.6%) and then lung metastasis was 19 cases (32.2%). In the 21 cases of multiple sites metastasis, triple sites was 6 cases (28.6%) and double sites was 15 cases (71.4%). PIK3CA mutation was the most frequent gene variation in all patients (34.5% of no metastasis group and 27.1% of metastasis group) but there was no difference between two groups(p-value = 0.278). BRCA 1 loss of function and BRCA2 loss of function were more frequent in metastasis group than no metastasis group(p-value = 0.033 and p-value = 0.024, respectively) but total counts was too small. We assessed the area under the curve (AUC) of the receiver operating characteristic (ROC) curve for predictive value. The AUC of ROC curve was 1.000 and also accuracy, precision, recall were 1.000. In addition, we conducted internal validation using 83 patients during 2015. When we applied a 0.5 threshold value with our predictive model, true negative was 81 and true positive was 2 among 83 patients. Finally, the accuracy of validation was 1.000. Conclusion Our predicted model could represent a useful and easy-to-access tool for the selection of patients with distant metastasis. After additional evaluation with large data and external validation, worldwide use of our model could be expected. Citation Format: Kim I, Choi HJ, Ryu JM, Lee SK, Yu JH, Kim SW, Nam SJ, Seo SW, Lee JE. A predictive model for distant metastasis in breast cancer patients using machine learning [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P2-08-52.

Bringing big data analytics closer to practice: A methodological explanation and demonstration of classification algorithms

BackgroundBig data analytics are becoming more prevalent due to the recent availability of health data. Yet in spite of evidence supporting the potential contribution of big data analytics to health policy makers and care providers, these tools are still too complex to be routinely used. Further, access to comprehensive datasets required for more accurate results is complex and costly. Consequently, big data analytics are mostly used by researchers and experts who are far removed from actual clinical practice. Hence, policy makers should allocate resources to encourage studies that clarify and simplify big data analytics so it can be used by non-experts (e.g., clinicians, practitioners and decision-makers who may not have advanced computer skills). It is also important to fund data collection and integration from various health IT, a pre-condition for any big data analytics project. ObjectivesTo methodologically clarify the rationale and logic behind several analytics algorithms to help non-expert users employ big data analytics by understanding how to implement relatively easy to use platforms as Azure ML. MethodsWe demonstrate the predictive power of four known algorithms and compare their accuracy in predicting early mortality of Congestive Heart Failure (CHF) patients. ResultsThe results of our models outperform those reported in the literature, attesting to the strength of some of the models, and the utility of comprehensive data. ConclusionsThe results support our call to policy makers to allocate resources to establishing comprehensive, integrated health IT systems, and to projects aimed at simplifying ML analytics.

Predictive Analysis of Financial Fraud Detection using Azure and Spark ML

This paper aims at providing valuable insights on Financial Fraud Detection on a mobile money transactional activity. We have predicted and classified the transaction as normal or fraud with a small sample and massive data set using Azure and Spark ML, which are traditional systems and Big Data respectively. Experimenting with sample dataset in Azure, we found that the Decision Forest model is the most accurate to proceed in terms of the recall value. For the massive data set using Spark ML, it is found that the Random Forest classifier algorithm of the classification model proves to be the best algorithm. It is presented that the Spark cluster gets much faster to build and evaluate models as adding more servers to the cluster with the same accuracy, which proves that the large scale data set can be predictable using Big Data platform. Finally, we reached a recall score with 0.73, which implies a satisfying prediction quality in predicting fraudulent transactions.

The Research on Stability of the Russian Banking System by Machine Learning Methods

The problem of stability of the Russian banking system is investigated. To describe the state of a commercial bank, we use a system of indicators, proposed by F.T. Aleskerov and his colleagues. For predicting the development of banking system, the machine learning system implemented in the Azure ML is used. To optimize the work of this software, it is suggested to use integral indicators.

A predictive model for high/low risk group according to oncotype DX recurrence score using machine learning

BackgroundOncotype DX(ODX) is a 21-gene breast cancer recurrence score(RS) assay that aids in decision-making for chemotherapy in early-stage hormone receptor-positive(HR+)breast cancer. We developed a prediction tool using machine learning for high- or low-risk ODX criteria (i.e., RS < 11 for low-risk; RS > 25 for high-risk). MethodsWe performed a retrospective review of 301 breast cancer patients who underwent surgery between April 2011 and July 2017 and then an ODX test at Samsung Medical Center in Seoul, Korea.Among them, 208 cases were defined as the modeling group and 76 cases were defined as the validation group. We built a supervised machine learning classification model using the Azure ML platform. ResultsFor the high RS group, accuracy was 0.903 through Two-class Decision Jungle method in test set. For the low RS group, the accuracy was 0.726 when the Two-class Neural Network method was applied. The AUC of the ROC curve was 0.917 in the high RS group and 0.744 in the low RS group in test set. In addition, we conducted an internal validation using 76 patients who underwent ODX testing between January 2017 and July 2017. The accuracy of validation was 0.880 in the high RS group and 0.790 in the low RS group. ConclusionWe developed a predictive model using machine learning that could represent a useful and easy-to-access tool for the selection of high ODX RS patients. After additional evaluation with large data and external validation, worldwide use of our model could be expected.

Sentiment Analysis Algorithms through Azure Machine Learning: Analysis and Comparison

The Sentimental Analysis (SA) is a widely known and used technique in the natural language processing realm. It is often used in determining the sentiment of a text. It can be used to perform social media analytics. This study sought to compare two algorithms; Logistic Regression, and Support Vector Machine (SVM) using Microsoft Azure Machine Learning. This was demonstrated by performing a series of experiments on three Twitter datasets (TD). Accordingly, data was sourced from Twitter a microblogging platform. Data were obtained in the form of individuals’ opinions, image, views, and twits from Twitter. Azure cloud-based sentiment analytics models were created based on the two algorithms. This work was extended with more in-depth analysis from another Master research conducted lately. Results confirmed that Microsoft Azure ML platform can be used to build effective SA models that can be used to perform data analytics.