Search

Turistički sektor može polučiti velike benefite od kvalitetne primjene analize podataka. Kvalitetan uvid u dinamiku rezervacija smještaja, potražnju i potrebe klijenata može ponuđaču turističkih usluga dati mogućnost donošenja poslovnih odluka koje će s jedne strane dodatno podići kvalitetu ponude i zadovoljstvo klijenata, a s druge rezultirati većim profitima. Zadatak ovog rada jest prikupiti podatkovni skup povezan uz turistički sektor, provesti eksploratornu analizu nad tim podacima i razviti prediktivni model za odabrane zavisne varijable. Konačno rješenje potrebno je realizirati u obliku programske skripte koje će implementirati razvijene modele i evaluirati njihovu učinkovitost uz komparativnu analizu korištenih metoda. The tourism sector can benefit from the quality application of data analysis. Quality insight into the dynamics of accommodation reservations, demand and customer needs can give the tourism service provider the opportunity to make business decisions that will on the one hand, further raise the quality of the offer and customer satisfaction, and on the other hand, result higher profits. The task of this paper is to collect a data set related to the tourism sector, to implement exploratory analysis of these data and develop a predictive model for selected dependent variables. The final solution needs to be implemented in the form of a program script that will implement the developed models and evaluate their effectiveness with comparative analysis of used methods.

Tema ovog rada je predviđanje bankrota tvrtke modelima strojnog učenja. Na početku je ukratko opisana povijest i podijela strojnog čcenja te problemi koji se rješavaju njime. Nakon toga, objašnjeni su modeli logističke regresije i naivnog bayesovog klasifikatora. Navedeni modeli koriste se kao implementacija iz alata Scikit-learn u sklopu praktičnog rada napisanog u jeziku Python. Napravljeno je i optimiranje parametara modela kako bi se poboljšale performanse. Rezultati modela prije i nakon optimizacije prikazani su grafički i tablično te je njihova usporedba napravljena na kraju 5. poglavlja. The topic of this paper is the prediction of company bankruptcy by machine learning models. At the beginning, the history and division of machine learning and the problems it solves are briefly described. After that, the models of logistic regression and the naive Bayesian classifier are explained. These models are used as an implementation of the Scikit-learn tool as part of a practical paper written in Python. Optimization of model parameters has also been done to improve performance. The results of the models before and after the optimization are presented graphically and tabularly and their comparison is made at the end of Chapter 5.

Cilj ovog rada je napraviti aplikaciju koju će korisnik koristiti kao pomoćni alat koji mu olakšava i ubrzava proces "oluje pisanja" (eng. Brainwriting) tako da automatizira pretragu web stranica s relevantnim informacijama, te integrira i prikazuje relevantne podatke na jednostavan način. Aplikacija se sastoji iz tri dijela koja čine cjelinu. Prvi dio je web stranica koja služi kao korisničko sučelje preko kojega korisnik može u slobodnoj formi unijeti zapise "oluje pisanja", te kasnije vidjeti rezultate pretrage sustava. Web stranica isto tako omogućuje korisniku da unosi svoje slike u sustava, te ih kasnije pregledava. Drugi dio aplikacije je model strojnog učenja koji na temelju zapisa "oluje pisanja", koju je unio korisnik, radi predviđanje filmskog žanra koji bi najbolje odgovarao tom zapisu. Treći dio sustava se odnosi na programe (botove) koji na temelju dobivenog filmskog žanra pretražuju relevantne izvore podataka na internetu kako bi prinašao pet najpopularnijih filmova za taj žanr. Nakon toga botovi pronalaze i dohvaćaju slike, citate i sažetak radnje za svaki od tih finova. Goal of this paper was to make web application that is used as a tool that will help speed up and automate process of acquiring relevant image templates based on user brainwriting input. After searching for images, quotes and plot summaries on relevant websites, our application integrates all this data and displays it in user friendly way. This application has three main components that work together as a whole. First component is website that enables users to enter their brainwriting input trough user interface and later displays search results in form of images and text. Second part of application is machine learning model that takes in as input brainwriting text and returns film genre that best fits users input. Third part of application is made of internet bots that, based on film genre, simulate (automate) users search the internet for five most popular films for given genre and then extract relevant data about them.

Reducing green urban infrastructure pose a huge threat to cities sustainability. It is important to monitor and track health of vegetation. For efficient planning of urban development and maintenance of green urban infrastructure key is to have up to date maps. Using satellite imagery is the easiest way to cover large city areas. In order to map vegetation, there are many possible solutions, however most effective way is using machine learning methods. Machine learning is divided into supervised and unsupervised classification and each can be divided into several different methods. Many authors have considered different methods and they use them to access accuracy on satellite image information extraction. They have used different satellite images and naturally higher resolution imagery results in better classification. However, there is still lack of comprehensive analysis of more supervised machine learning methods on similar cities. This paper aims to provide analysis of four different machine learning methods: support vector machine, artificial neural network, naïve Bayes and random forest. The objective of the support vector machine algorithm is to find a hyperplane in an N-dimensional space that distinctly classifies the data points where hyperplanes are decision boundaries that help classify the data points and support vectors data points that are closer to the hyperplane and influence the position and orientation of the hyperplane. Artificial neural network are brain-inspired systems which are intended to replicate the way that humans learn. Neural networks consist of input and output layers, and hidden layers. They are excellent tools for finding patterns which are far too complex or numerous for a human programmer to extract and teach the machine to recognize. A Naive Bayes classifier is a probabilistic machine learning model that’s used for classification task and the crux of the classifier is based on the Bayes theorem. Random Forest creates a forest and makes it random and is an ensemble of Decision Trees, most of the time trained with the bagging method which general idea is that a combination of learning models increases the overall result. All of the mentioned methods will be tested on Sentinel-2A imagery. Sentinel-2A multispectral imager has thirteen sensors which is useful in vegetation extraction. Methods will be compared using error matrix and kappa statistics.

Sustavi za digitalno učenje generiraju ogromne količine podataka iz kojih se dubinskom analizom mogu izvući informacije koje nisu dostupne jednostavnim pretraživanjem. Primjenom programskog okruženja Python razvijen je sustav za automatsku klasifikaciju zadataka predmeta Osnove elektrotehnike Fakulteta elektrotehnike i računarstva u Zagrebu po prethodno dodijeljenim glavnim i sporednim konceptima. Objašnjene su korištene metode predobrade i vektorizacije zadataka. Klasifikacija je izvedena po glavnim i sporednim konceptima odvojeno, združenim glavnim i sporednim konceptima i hijerarhijski. Izgrađeno je pet različitih klasifikatora te su komentirani dobiveni rezultati za svaki slučaj klasifikacije. Digital learning systems generate huge amounts of data from which information that is not accessible by simple search can be collected by applying data mining methods. Using the Python program environment, an automated classification system for the exam questions of Electrical Engineering course at the Faculty of Electrical Engineering and Computing in Zagreb was developed over previously assigned primary and secondary concepts. Used pre-processing and vectorization methods are explained. Classification is performed by primary and secondary concepts separately, joined primary and secondary concepts, and hierarchically. Five different classifiers were constructed and the results for each case were commented.

Cilj je usporediti algoritme strojnog učenja koji rješavaju problem klasifikacije na različitim skupovima podatka. Usporedba se temelji na točnosti predviđanja algoritama na njima prije neviđenim podacima. Algoritmi koji se uspoređuju su: Naivni Bayesov klasifikator, Stablo odluke, Najbliži susjed, Linearna diskriminantna analiza, Logistička regresija i Stroj potpornih vektora. Usporedba se provodila nad tri različita skupa podataka. Rezultati usporedbe su pokazali da algoritmi Stablo odluke i Najbliži susjed imaju dobru točnost neovisno o vrsti podataka, dok algoritmi Linearne diskriminantne analize, Logističke regresije i Stroja potpornih vektora imaju dobre i bolje rezultate za specifičnije vrste podatka. The goal is to compare different machine learning algorithms for classification problems using different datasets. The comparison is based on predicting the classes of data not previously seen by the algorithms. Algorithms used in the comparison are: Naive Bayes classifier, Decision tree, Nearest neighbour, Linear discriminant analysis, Logistic regression, Support vector machine. Three different datasets were used for the comparison. Results of the comparison show that accuracies of Nearest neighbour and Decision tree algorithms are good, no matter the data type, while algorithms like Linear discriminant analysis, Logistic regression and Support vector machine have better result for specific data types.

S ciljem približavanja baza podataka ljudima bez znanja potrebnih za upravljanje istima počela su se razvijati jezična sučelja bazama podataka. Kako bi se izgradilo stabilno jezično sučelje nužno je odbacivati loše upite. Taj problem riješen je u ovom radu uporabom metoda otkrivanja novih vrijednosti i strojnog učenja. Sljedeći korak pri izgradnji jezičnog sučelja je predviđanje vrste pitanja u upitu, kako bi se upit mogao ispravno prevesti u jezik upita. U opsegu ovog rada taj problem riješen je klasifikacijom u 14 klasa koristeći metode obrade prirodnog jezika i strojnog učenja. U radu su detaljno opisani algoritmi koji su korišteni te postupci kojima su se rezultati dobili. Skup podataka na kojima su modeli učeni ručno je izgrađen te označen. Natural language interfaces to databases began to develop with the goal to bring databases closer to the people with no knowledge about them. To make a natural language interface it is necessary to filter out spam queries. That problem is solved in this thesis by using methods of novelty detection and machine learning. The next step in creating a natural language interface is predicting the question type in query, in order for the query to be easily translated to query language. In the scope of this thesis that problem was solved by classification into 14 classes using methods of natural language processing and machine learning. Algorithms that were used in the thesis are also thoroughly explained as well as the methods by which we got results. Dataset on which the models were trained was compiled and labeled by hand.

Zbog postojanja vrlo velike količine tekstualnih dokumenata potrebno ih je, radi brže, jeftinije i lakše obrade, organizirati, odnosno podijeliti prema zajedničkim svojstvima na zadane klase. Takav postupak nazivamo klasifikacija koja je, osim na tekstualne dokumente, primjenjiva i na druge vrste podataka. S obzirom na svojstva podataka koje organiziramo, potrebno je odabrati onu metodu klasifikacije koja će najbolje odrediti pripadnost klasi. Odluka se temelji na teorijskim saznanjima o pojedinim metodama i usporedbi dobivenih rezultata. Moguće je da iste metode, primijenjene na različite skupove podataka, rezultiraju vrlo različitom točnosti određivanja klasa. U radu su opisana dva klasifikatora koji pripadaju različitim vrstama strojnog učenja. To su klasifikator najbližih susjeda i naivni Bayesov klasifikator. Za primjenu klasifikacije moguće je koristiti Apache Lucene, programsku knjižnicu otvorenog koda. In order to efficiently process large number of textual documents it is necessary to organize them by assigning a class for each document. Because of that we use classification methods which can be applied to other types of data as well. Considering the properties of documents set, it is necessary to use a classification method which will result with the best classification result. The decision has to be made based on theoretical knowledge and regarding the given set of documents because the quality of the same method can vary regarding the properties of given set of documents. Two classification methods with different properties are described in this thesis. These are \textit{k}-Nearest Neighbor classification and Naive Bayes classifier. Apache Lucene, an open source information retrieval library, can be used for implementing text classification.