BazTech - Yadda

1

Application of sample advisory systems in medicine

Woehl Agnieszka, Zapotoczny Kacper, Żaba Julia, Nagi Łukasz

Sustainable Production, Instrumentation and Engineering Sciences

|

2023

|

vol. 1, nr 1

13--17

EN

Artificial intelligence is a field that has been rapidly developing in various areas of knowledge in recent years. Its application in medicine can support the intensive development of research in health care and improve and ac-celerate the operation of many medical facilities. This article presents sev-eral examples of expert systems that can find application in diagnosing and preparing a patient for selected tests. Expert systems can also find appli-cation in the rapid selection of rehabilitation, medical or support equip-ment and devices with which medical facilities are supplied. In this article, the reader will also find a sample application that will perform this func-tion. The article presents the elements of which a correct expert system should consist. For each application, tests have been carried out to show the correctness of the system. The purpose of the article was to show the capabilities of the expert system and its application in medical fields.

2

Designing Smart Antennas Using Machine Learning Algorithms

Samantaray Barsa, Das Kunal Kumar, Roy Jibendu Sekhar

Journal of Telecommunications and Information Technology

|

2023

|

nr 4

46--52

EN

Smart antenna technologies improve spectral efficiency, security, energy efficiency, and overall service quality in cellular networks by utilizing signal processing algorithms that provide radiation beams to users while producing nulls for interferers. In this paper, the performance of such ML solutions as the support vector machine (SVM) algorithm, the artificial neural network (ANN), the ensemble algorithm (EA), and the decision tree (DT) algorithm used for forming the beam of smart antennas are compared. A smart antenna array made up of 10 half-wave dipoles is considered. The ANN method is better than the remaining approaches when it comes to achieving beam and null directions, whereas EA offers better performance in terms of reducing the side lobe level (SLL). The maximum SLL is achieved using EA for all the user directions. The performance of the ANN algorithm in terms of forming the beam of a smart antenna is also compared with that of the variable-step size adaptive algorithm.

3

Prediction of Selected Mechanical Properties in Austempered Ductile Iron with Different Wall Thickness by the Decision Support Systems

Jaśkowiec Krzysztof, Opaliński Andrzej, Kustra Piotr, Jach D., Wilk-Kołodziejczyk Dorota

Archives of Foundry Engineering

|

2023

|

Vol. 23, iss. 2

137--144

EN

The structure of Austempered Ductile Iron (ADI) is depend of many factors at individual stages of casting production. There is a rich literature documenting research on the relationship between heat treatment and the resulting microstructure of cast alloy. A significant amount of research is conducted towards the use of IT tools for indications production parameters for thin-walled castings, allowing for the selection of selected process parameters in order to obtain the expected properties. At the same time, the selection of these parameters should make it possible to obtain as few defects as possible. The input parameters of the solver is chemical composition Determined by the previous system module. Target wall thickness and HB of the product determined by the user. The method used to implement the solver is the method of Particle Swarm Optimization (PSO). The developed IT tool was used to determine the parameters of heat treatment, which will ensure obtaining the expected value for hardness. In the first stage, the ADI cast iron heat treatment parameters proposed by the expert were used, in the next part of the experiment, the settings proposed by the system were used. Used of the proposed IT tool, it was possible to reduce the number of deficiencies by 3%. The use of the solver in the case of castings with a wall thickness of 25 mm and 41 mm allowed to indication of process parameters allowing to obtain minimum mechanical properties in accordance with the PN-EN 1564:2012 standard. The results obtained by the solver for the selected parameters were verified. The indicated parameters were used to conduct experimental research. The tests obtained as a result of the physical experiment are convergent with the data from the solver.

4

Data-driven fault detection and diagnosis for centralised chilled water air conditioning system

Sulaiman Noor Asyikin, Chuink Kai Wern, Zainudin Muhammad Noorazlan Shah, Yusop Azdiana Md, Sulaiman Siti Fatimah, Abdullah Md Pauzi

Przegląd Elektrotechniczny

|

2022

|

R. 98, nr 1

217--221

EN

The air conditioning system is complex and consumes the most energy in the building. Due to its complexity, it is difficult to identify faults in the system immediately. In this project, fault detection and diagnosis system using decision tree classifier model was developed to detect and diagnose faults in a chilled water air conditioning system. The developed model successfully classified normal condition and five common faults for more than 99% accuracy and precision. A graphical user interface of the system was also developed to ease the users.

PL

System klimatyzacji jest złożony i zużywa najwięcej energii w budynku. Ze względu na swoją złożoność trudno jest od razu zidentyfikować usterki w systemie. W ramach tego projektu opracowano system wykrywania i diagnostyki usterek wykorzystujący model klasyfikatora drzewa decyzyjnego do wykrywania i diagnozowania usterek w systemie klimatyzacji wody lodowej. Opracowany model pomyślnie sklasyfikował stan normalny i pięć typowych usterek, zapewniając ponad 99% dokładności i precyzji. W celu ułatwienia użytkownikom opracowano również graficzny interfejs użytkownika systemu.

5

Features of Creating a System of Space Monitoring of Water-Supplied Territories for Irrigation in the South of Kazakhstan

Malakhov Dmitry V., Tskhay Mikhail, Kalashnikov Alexander A., Bekmukhamedov Nurlan E., Kalashnikov Pavel A., Baizakova Aigul

Journal of Ecological Engineering

|

2022

|

Vol. 23, nr 11

202--216

EN

The location of a significant part of the agricultural territories of Kazakhstan in the risk agriculture zone implies the development and further application of an objective monitoring system for irrigated territories. The purpose of the study was to develop methods for on-the-spot and long-term recognition of irrigated massifs and verification of methods in the conditions of the territories of southern Kazakhstan. The paper describes the methods of on-the-spot recognition of irrigated fields, the general assessment of irrigated areas for the growing season, as well as the method of recognizing promising areas for irrigation. The on-the-spot recognition of the fields is based on the use of such spectral indices as the Global Vegetation Moisture Index, Green Normalized Difference Vegetation Index, Normalized Difference Vegetation Index, and the xanthophyll index, combined into a single system by the Decision Tree algorithm. The assessment of irrigated areas is based on differences in the physiological state of plants in conditions of normal water supply and plants experiencing a lack of moisture. The evaluation system includes the calculation of the temperature difference according to the corresponding satellite data and the calculation of the difference in vegetation indices for the same period. The difference in vegetation indices in irrigated fields has positive values due to a steady increase in green biomass, and the temperature difference, on the contrary, is negative or zero, since healthy plants, with normal water supply, actively evaporate moisture to maintain optimal temperatures of biochemical processes. To develop these methods, ground data from 2017–2021 were used. Verification of the methods with ground data demonstrated acceptable accuracy (87% in the on-the-spot assessment of irrigated fields; 60–90% in the general assessment of irrigated areas), while the methods have significant potential for further improvement.

6

Diagnostics of the drive shaft bearing based on vibrations in the high-frequency range as a part of the vehicle's self-diagnostic system

Nowakowski Tomasz, Komorski Pawel

Eksploatacja i Niezawodność

|

2022

|

Vol. 24, no. 1

70--79

EN

Currently, one of the trends in the automotive industry is to make vehicles as autonomous as possible. In particular, this concerns the implementation of complex and innovative selfdiagnostic systems for cars. This paper proposes a new diagnostic algorithm that evaluates the performance of the drive shaft bearings of a road vehicle during use. The diagnostic parameter was selected based on vibration measurements and machine learning analysis results. The analyses included the use of more than a dozen time domain features of vibration signal in different frequency ranges. Upper limit values and down limit values of the diagnostic parameter were determined, based on which the vehicle user will receive information about impending wear and total bearing damage. Additionally, statistical verification of the developed model and validation of the results were performed.

7

Rule modeling of ADI cast iron structure for contradictory data

Soroczyński Artur, Biernacki Robert, Kochański Andrzej

Computer Methods in Materials Science

|

2022

|

Vol. 22, No. 4

211--228

EN

Ductile iron is a material that is very sensitive to the conditions of crystallization. Due to this fact, the data on the cast iron properties obtained in tests are significantly different and thus sets containing data from samples are contradictory, i.e. they contain inconsistent observations in which, for the same set of input data, the output values are significantly different. The aim of this work is to try to determine the possibility of building rule models in conditions of significant data uncertainty. The paper attempts to determine the impact of the presence of contradictory data in a data set on the results of process modeling with the use of rule-based methods. The study used the well-known dataset (Materials Algorithms Project Data Library, n.d.) pertaining to retained austenite volume fraction in austempered ductile cast iron. Two methods of rulebased modeling were used to model the volume of the retained austenite: the decision trees algorithm (DT) and the rough sets algorithm (RST). The paper demonstrates that the number of inconsistent observations depends on the adopted data discretization criteria. The influence of contradictory data on the generation of rules in both algorithms is considered, and the problems that can be generated by contradictory data used in rule modeling are indicated.

8

Evaluation of the effect of various surficial pollutants and environmental condition on surface friction performance of road pavement

Khabiri Mohammad Mehdi, Afkhamy-Meybodi Pooya, Montazeri Ali Mohammad

Zeszyty Naukowe. Transport / Politechnika Śląska

|

2021

|

z. 112

99--111

EN

Skip resistance of asphalt is an important parameter that can influence the safety of drivers on roads. Although there is a linear relationship between slipping on road surfaces and accidents, the impacts of pollutants for decreasing friction of roads is clear to researchers. Moisture and temperature influence friction and safety. In this research in SMA samples, three different gradations with the maximum nominal sizes of 19, 12.5 and 9.5, based on international standards were used. For polluting the surface, five materials that are found on roads were used, including fine-grained soil, sand, oil, soot and rubber powder. To measure the skip resistance, the British pendulum tester was used and for analysing macro-texture, the sand patch method was used. The results of this research showed that by increasing the maximum nominal size of aggregates, the depth of macro-texture in surfaces are grown and this is due to the decrease of fine aggregates in larger gradations. Because of the higher flexibility of pure bitumen, the applied compression pressure on rigid aggregates can cause indentations in the substrate and result in declining the roughness height of aggregates in the mixed surface. This leads to declining the hysteresis part of friction by increasing temperature.

9

Prediction of Post-Diagnostic Decisions for Tested Hand Grenades’ Fuzes Using Decision Trees

Ampuła Dariusz

Problemy Mechatroniki : uzbrojenie, lotnictwo, inżynieria bezpieczeństwa

|

2021

|

Vol. 12, Nr 2 (44)

39--54

EN

The article presents a brief history of creation of decision trees and defines the purpose of the undertaken works. The process of building a classification tree, according to the CHAID method, is shown paying particular attention to the disadvantages, advantages, and characteristics features of this method, as well as to the formal requirements that are necessary to build this model. The tree’s building method for UZRGM (Universal Modernised Fuze of Hand Grenades) fuzes was characterized, specifying the features of the tested hand grenade fuzes and the predictors used that are necessary to create the correct tree model. A classification tree was built basing on the test results, assuming the accepted post-diagnostic decision as a qualitative dependent variable. A schema of the designed tree for the first diagnostic tests, its full structure and the size of individual classes of the node are shown. The matrix of incorrect classifications was determined, which determines the accuracy of incorrect predictions, i.e., correctness of the performed classification. A sheet with risk assessment and standard error for the learning sample and the v-fold cross-check were presented. On the selected examples, the quality of the resulting predictive model was assessed by means of a graph of the cumulative value of the lift coefficient and the "ROC" curve.

PL

We artykule przedstawiono krótką historię powstania drzew decyzyjnych oraz określono cel podjętych prac. Pokazano proces budowy drzewa klasyfikacyjnego według metody CHAID, zwracając szczególną uwagę na wady, zalety oraz cechy charakterystyczne tej metody a także na wymagania formalne, które są niezbędne do zbudowania tego modelu. Scharakteryzowano metodę budowy drzewa dla zapalników UZRGM, określając cechy badanych zapalników do granatów ręcznych oraz zastosowane predyktory, które są konieczne do tworzenia prawidłowego modelu drzewa. Zbudowano drzewo klasyfikacyjne na podstawie posiadanych wyników badań, przyjmując jako jakościową zmienną zależną przyjętą decyzję podiagnostyczną. Pokazano schemat zaprojektowanego drzewa dla pierwszych badań diagnostycznych, jego pełną strukturę oraz liczności poszczególnych klas węzła. Określono macierz błędnych klasyfikacji, która określa trafność błędnych predykcji, czyli poprawność dokonanej klasyfikacji. Przedstawiono arkusz z oceną ryzyka oraz błędem standardowym dla próby uczącej i v-krotnego sprawdzianu krzyżowego. Na wybranych przykładach oceniono jakość powstałego modelu predykcyjnego za pomocą wykresu skumulowanej wartości współczynnika przyrostu oraz krzywej "ROC".

10

Induction of decision trees for building knowledge bases of production processes

Matuszny Marcin

Polish Technical Review

|

2021

|

No. 1

21--25

EN

The article presents the process of acquiring the knowledge based on the induction of decision trees, graphically illustrating the differences between acquiring the knowledge in a traditional way, from the expert, and the process of acquiring the knowledge supported by the machine learning methods. The methods of acquiring the knowledge are discussed and specified. The practical part represents the use of De Treex 4.0 software dedicated to the induction of decision trees, which is a part of the Sphinx 4.0 artificial intelligence package.

PL

W artykule przedstawiono proces pozyskiwania wiedzy w oparciu o indukcję drzew decyzyjnych, w sposób graficzny zilustrowano różnice pomiędzy pozyskiwaniem wiedzy w sposób tradycyjny, od eksperta, a także procesem pozyskiwania wiedzy wspomaganym metodami uczenia maszynowego. Omówiono i wyszczególniono metody pozyskiwania wiedzy. W części praktycznej przedstawiono wykorzystanie oprogramowania DeTreex 4.0 dedykowanego do indukcji drzew decyzyjnych wchodzącego w skład pakietu sztucznej inteligencji Sphinx 4.0.

11

Handwrittenword recognition using fuzzy matching degrees

Wróbel Michał, Starczewski Janusz T., Fijałkowska Justyna, Siwocha Agnieszka, Napoli Christian

Journal of Artificial Intelligence and Soft Computing Research

|

2021

|

Vol. 11, No. 3

229--242

EN

Handwritten text recognition systems interpret the scanned script images as text composed of letters. In this paper, efficient offline methods using fuzzy degrees, as well as interval fuzzy degrees of type-2, are proposed to recognize letters beforehand decomposed into strokes. For such strokes, the first stage methods are used to create a set of hypotheses as to whether a group of strokes matches letter or digit patterns. Subsequently, the second-stage methods are employed to select the most promising set of hypotheses with the use of fuzzy degrees. In a primary version of the second-stage system, standard fuzzy memberships are used to measure compatibility between strokes and character patterns. As an extension of the system thus created, interval type-2 fuzzy degrees are employed to perform a selection of hypotheses that fit multiple handwriting typefaces.

12

Optimal Ensemble Learning Based on Distinctive Feature Selection by Univariate ANOVA-F Statistics for IDS

Shakeela Shaikh, Sai Shankar N., Mohan Reddy P., Kavya Tulasi T., Mahesh Koneru M.

International Journal of Electronics and Telecommunications

|

2021

|

Vol. 67, No. 2

267--275

EN

Cyber-attacks are increasing day by day. The generation of data by the population of the world is immensely escalated. The advancements in technology, are intern leading to more chances of vulnerabilities to individual’s personal data. Across the world it became a very big challenge to bring down the threats to data security. These threats are not only targeting the user data and also destroying the whole network infrastructure in the local or global level, the attacks could be hardware or software. Central objective of this paper is to design an intrusion detection system using ensemble learning specifically Decision Trees with distinctive feature selection univariate ANOVA-F test. Decision Trees has been the most popular among ensemble learning methods and it also outperforms among the other classification algorithm in various aspects. With the essence of different feature selection techniques, the performance found to be increased more, and the detection outcome will be less prone to false classification. Analysis of Variance (ANOVA) with F-statistics computations could be a reasonable criterion to choose distinctives features in the given network traffic data. The mentioned technique is applied and tested on NSL KDD network dataset. Various performance measures like accuracy, precision, F-score and Cross Validation curve have drawn to justify the ability of the method.

13

Application of machine learning and rough set theory in lean maintenance decision support system development

Antosz Katarzyna, Jasiulewicz-Kaczmarek Małgorzata, Paśko Łukasz, Zhang Chao, Wang Shaoping

Eksploatacja i Niezawodność

|

2021

|

Vol. 23, no. 4

695--708

EN

Lean maintenance concept is crucial to increase the reliability and availability of maintenance equipment in the manufacturing companies. Due the elimination of losses in maintenance processes this concept reduce the number of unplanned downtime and unexpected failures, simultaneously influence a company’s operational and economic performance. Despite the widespread use of lean maintenance, there is no structured approach to support the choice of methods and tools used for the maintenance function improvement. Therefore, in this paper by using machine learning methods and rough set theory a new approach was proposed. This approach supports the decision makers in the selection of methods and tools for the effective implementation of Lean Maintenance.

14

Data Mining for Bankruptcy Prediction: An Experiment in Vietnam

Hung Dang Ngoc, Binh Vu Thi Thanh

Annals of Computer Science and Information Systems

|

2021

|

Vol. 28

175--184

EN

In the history of the world economy, the bankruptcy of some large companies has caused global financial crises. The study aimed to postulate a model of bankruptcy prediction for listed companies on Vietnam's stock market. The research used six popular algorithms in data mining to predict bankruptcy risk with data collected from 4693 observations in the period 2009-2020. The research results showed that Logistic algorithms, Artificial Neural Network, Decision Tree have a high level of predicting bankruptcy with an accuracy of 98%. The study identified the three most important indicators: inventory turnover ratio, debt to equity ratio, and debt ratio that affect the corporate bankruptcy prediction. The study showed the threshold points of 10-indicators to avoid bankruptcy likelihood. These results recommended that the model could be applied in practice to reduce risks for businesses and investors in the Vietnamese market.

15

Drought classification using gradient boosting decision tree

Danandeh Mehr Ali

Acta Geophysica

|

2021

|

Vol. 69, no. 3

909--918

EN

This paper compares the classification and prediction capabilities of decision tree (DT), genetic programming (GP), and gradient boosting decision tree (GBT) techniques for one-month ahead prediction of standardized precipitation index in Ankara province and standardized precipitation evaporation index in central Antalya region. The evolved models were developed based on multi-station prediction scenarios in which observed (reanalyzed) data from nearby stations (grid points) were used to predict drought conditions in a target location. To tackle the rare occurrence of extreme dry/wet conditions, the drought series at the target location was categorized into three classes of wet, normal, and dry events. The new models were trained and validated using the frst 70% and last 30% of the datasets, respectively. The results demonstrated the promising performance of GBT for meteorological drought classification. It provides better performance than DT and GP in Ankara; however, GP predictions for Antalya were more accurate in the testing period. The results also exhibited that the proposed GP model with a scaled sigmoid function at root can efortlessly classify and predict the number of dry, normal, and wet events in both case studies.

16

Artificial Intelligence Algorithms for the Analysis of Mechanical Property of Friction Stir Welded Joints by using Python Programming

Mishra Akshansh

Welding Technology Review

|

2020

|

R. 92, nr 6

7--16

EN

In modern computational science, the interplay existing between machine learning and optimization process marks the most vital developments. Optimization plays an important role in mechanical industries because it leads to reduce in material cost, time consumption and increase in production rate. The recent work focuses on performing the optimization task on Friction Stir Welding process for obtaining the maximum Ultimate Tensile Strength (UTS) of the friction stir welded joints. Two machine learning algorithms i.e. Artificial Neural Network (ANN) and Decision Trees regression model are selected for the purpose. The input variables are Tool Rotational Speed (RPM), Tool Traverse Speed (mm/min) and Axial Force (KN) while the output variable is Ultimate Tensile Strength (MPa). It is observed that in case of the Artificial Neural Networks the Root Mean Square Errors for training and testing sets are 0.842 and 0.808 respectively while in case of Decision Trees regression model, the training and testing sets result Root Mean Square Errors of 11.72 and 14.61. So, it can be concluded that ANN algorithm gives better and accurate result than Decision Tree regression algorithm.

PL

We współczesnych obliczeniach naukowych wzajemna zależność między uczeniem maszynowym a procesem optymalizacji wyznacza najbardziej istotne osiągnięcia. Optymalizacja odgrywa ważną rolę w przemyśle mechanicznym, ponieważ prowadzi do obniżenia kosztów materiałów, zużycia czasu i wzrostu szybkości produkcji. Ostatnie prace skupiają się na wykonaniu optymalizacji procesu zgrzewania tarciowego z przemieszaniem w celu uzyskania maksymalnej wytrzymałości na rozciąganie (UTS) połączeń zgrzewanych tarciowo z przemieszaniem. Do tego celu wybrano dwa algorytmy uczenia maszynowego, tj. Sztuczną sieć neuronową (ANN) i model decyzyjnego drzewa regresyjnego. Zmienne wejściowe to prędkość obrotowa narzędzia [obr/min], prędkość posuwu narzędzia [mm/min] i siła osiowa [kN], natomiast zmienną wyjściową jest maksymalna wytrzymałość na rozciąganie [MPa]. Zaobserwowano, że w przypadku sztucznych sieci neuronowych średnie błędy kwadratowe zbiorów uczących i testowych wynoszą odpowiednio 0,842 i 0,808, podczas gdy w przypadku modelu decyzyjnego drzewa regresji zbiory uczące i testujące dają średnie błędy kwadratowe 11,72 i 14,61. Można więc stwierdzić, że algorytm ANN daje lepsze i dokładniejsze wyniki niż algorytm regresji drzewa decyzyjnego.

17

Building decision trees based on production knowledge as support in decision-making process

Matuszny Marcin

Production Engineering Archives

|

2020

|

Vol. 26, Iss. 2

36--40

EN

The article presents sources of production knowledge and thoroughly describes its identification which on the construction of decision trees, and on the construction of knowledge bases for production processes. The problems that arise during the technical preparation of production are briefly characterized and the advanced algorithm with which decision trees can be built is described in detail. A decision tree was built based on real data from the manufacturing company. Decision trees are presented as a method of knowledge representation.

18

Utilizing decision trees on employee decision-making processes: a model proposal

Yüncü Volkan, Fidan Üzeyir

Organizacja i Zarządzanie : kwartalnik naukowy

|

2020

|

nr 3

137--152

EN

Introduction/background: This paper offers an idiosyncratic relational framework built on the organizational silence theory and the organizational support theory. It exploits the distinct advantages that using decision trees in classification and prediction applications offer to form a unique predictive model. Aim of the paper: This paper argues that a relational framework built on the organizational silence theory and the organizational support theory can give important clues about how employees make certain decisions in the workplace as well as about factors that have an impact on their decision-making processes. Materials and methods: The research applies decision trees learning – a data mining technique – to unfold the hidden patterns and unprecedented relationships between the two constructs that until now had not been revealed. Results and conclusions: The suggested model, which consists of rules, exhibits the effects of perceived organizational support and employee silence behavior on employee decisions with an approximately 79% correct classification rate, showing the success of the model as well as its appropriate relational framework. The presented findings indicate that a relational framework built on the organizational silence theory and the organizational support theory has a lot to offer in terms of building effective HR strategies and policies. The study also extends the understanding of the antecedents of silence behavior in different social contexts.

19

Decision trees in the tests of artillery igniters

Ampuła Dariusz

Journal of KONBiN

|

2020

|

Vol. 50, iss. 1

95--116

EN

The article addressed the method for building decision trees paying attention to the binary character of the tree structure. The methodology for building our decision tree for KW-4 igniters was presented. It involves determining features of tested igniters and applied predictors, which are necessary to create the correct model of the tree. The classification tree was built based on the possessed test results, determining the adopted post-diagnostic decision as the qualitative independent variable. The schema of the resultant classification tree and the full structure of this tree together with the results in end nodes were shown. The obtained graphic and tabular sequence of the designed tree was characterized, and the prediction accuracy was evaluated on the basis of the resultant matrix of incorrect classifications. The quality of the resultant predictive model was assessed on the basis of the chosen examples by means of the 'ROC' curve and the graph of the cumulative value of increase coefficient.

PL

W artykule opisano metodę budowy drzew decyzyjnych zwracając uwagę na binarny charakter struktury drzewa. Przedstawiono metodykę budowy drzewa dla zapłonników typu KW-4, określając cechy badanych zapłonników oraz zastosowane predyktory, które są niezbędne do tworzenia prawidłowego modelu drzewa. Na podstawie posiadanych wyników badań, zbudowano drzewo klasyfikacyjne, określając jako jakościową zmienną niezależną przyjętą decyzję podiagnostyczną. Pokazano schemat powstałego drzewa klasyfikacyjnego oraz pełną strukturę tego drzewa łącznie z wynikami w węzłach końcowych. Scharakteryzowano uzyskaną graficzną i tabelaryczną sekwencję zaprojektowanego drzewa oraz oceniono trafność predykcji na podstawie powstałej macierzy błędnych klasyfikacji. Oceniono na wybranych przykładach jakość powstałego modelu predykcyjnego za pomocą krzywej „ROC” oraz wykresu skumulowanej wartości współczynnika przyrostu.

20

Eksploracja danych i uczenie maszynowe w aspektach pozyskiwania wiedzy o produkcji i przetwarzaniu metali na potrzeby Przemysłu 4.0

Regulski Krzysztof

Hutnik, Wiadomości Hutnicze

|

2020

|

Vol. 87, nr 4

75--81

PL

Metodyka automatycznego odkrywania wiedzy o procesach wytwarzania i przetwarzania metali obejmuje problemy związane z (1) akwizycją danych i integracją ich w aspekcie dalszej eksploracji, (2) doborem i adaptacją metod uczenia maszynowego ― indukcji reguł, predykcji zmiennych ilościo wych i jakościowych, (3) formalizacją wiedzy w odpowiednich reprezentacjach: regułowej, zbiorów rozmytych, zbiorów przybliżonych czy wreszcie logiki deskrypcyjnej oraz (4) integracją wiedzy w repozytoriach opisanych modelami semantycznymi, czyli ontologiami. Autor przedstawił możliwość osiągnięcia równowagi pomiędzy wygodą użytkowania a precyzją w przypadku pozyskiwania wiedzy z małych zbiorów. Badania wykazały, że drzewa decyzyjne są wygodnym narzędziem odkrywania wiedzy i dobrze radzą sobie z problemami silnie nieliniowymi, a wprowadzenie dyskretyzacji poprawia ich działanie. Zastosowanie metod analizy skupień umożliwiło też wyciąganie bardziej ogólnych wniosków, przez co udowodniono tezę, że granulacja informacji pozwala znaleźć wzorce nawet w małych zbiorach danych. Opracowano w ramach badań procedurę postępowania w analizie małych zbiorów danych eksperymentalnych dla modeli multistage, multivariate & multivariable, co może w znacznym stopniu uprościć takie badania w przyszłości.

EN

The methodology of automatic knowledge discovery about metal production and processing processes includes problems related to (1) data acquisition and integration in the aspect of further exploration, (2) selection and adaptation of machine learning methods - rule induction, quantitative and qualitative variable prediction, (3) formalization knowledge in appropriate representations: rule, fuzzy sets, rough sets and finally descriptive logic, and (4) integration of knowledge in repositories described by semantic models or ontologies. The author presented the possibility of achieving a balance between ease of use and precision when acquiring knowledge from small collections. Research has shown that decision trees are a convenient tool for discovering knowledge and that they deal well with strongly non-linear problems, and the introduction of discretization improves their operation. The use of cluster analysis methods also made it possible to draw more general conclusions, which proved the thesis that granulation of information allows finding patterns even in small data sets. As part of the research, a procedure was developed for analyzing small experimental data sets for multistage, multivariate & multivariable models, which can greatly simplify such research in the future.