1. Napovedovanje lastništva podjetij na osnovi analize omrežij družbenikov
- Author
-
Jončevski, Igor and Bajec, Marko
- Subjects
computer and information science ,računalništvo ,napovedni model ,computer science ,statistika ,data mining ,graph ,univerzitetni študij ,prediction model ,pomembnost vozlišč ,obdelava podatkov ,računalništvo in informatika ,diploma ,statistics ,diplomske naloge ,napovedovanje podatkov ,network ,data prediction ,udc:004.9(043.2) ,node importance ,graf ,podatkovno rudarjenje ,data processing ,omrežje - Abstract
Namen diplomskega dela je bil izdelava napovednega modela, s katerim bi se lahko napovedovala sprememba lastništva v podjetju. Model je rezultat analize omrežja družbenikov, ki so zgrajena na podatkih o družbenikih in njihovi prisotnosti v podjetjih. Lastništvo je v bistvu stopnja vozlišča v omrežjih, ki so bila analizirana. V procesu grajenja modela je bistvenega pomena obdelava podatkov in njihova predstavitev v podobi omrežja. To je bilo narejeno s pomočjo programskega jezika Java in programskih vmesnikov OWL in Gephi. Dobljena omrežja v formi grafov je bilo treba naprej analizirati. Nadaljnja analiza je poskrbela za pridobitev mere pomembnosti vozlišč v omrežju, ker je bila njihova uporaba ključnega pomena v procesu napovedovanja. Dobljene mere so osnova za različne statistične metode, metode podatkovnega rudarjenja in metode strojnega učenja. Ugotovitve, pridobljene iz teh metod so pripeljale korak bližje do zamišljenega modela. Končni rezultat je napovedni model, ki je lahko osnova za namizno ali spletno aplikacijo, ki bi lahko služila za napovedovanje, ne le spremembe lastništva, ampak tudi drugih podatkov. The purpose of this thesis was the realization of a prediction model, with which we could predict the change of ownership in a network. The prediction model is a result of the analysis process of board collaboration networks, where the networks are in fact a representation of data for stockholders and their presence in a certain company. The ownership is represented with a node's degree in the networks which were analyzed. In the model realization process, data processing and their proper representation in the form of a network is essential. For that purpose, we used the Java programming language, coupled with the Application programming interfaces (APIs) OWL and Gephi. The resulting networks, represented as graphs, needed to be further analyzed in order for us to acquire the node importance metrics in the network, which were crucial for the prediction process. The acquired metrics are the basis for various statistical, data mining and machine learning methods. The results of those methods lead us to the creation of the model that we imagined in the first place. The end result can be the basis for a desktop or a web application who could predict not just ownership change, but also other data.
- Published
- 2015