Search

Objective: Informational social support is one of the main reasons for patients to visit online health communities (OHCs). Calls have been made to investigate the objective quality of such support in the light of a worrying number of inaccurate online health-related information. The main aim of this study is to conceptualize the Quality of Informational Social Support (QISS) and develop and test a measure of QISS for content analysis. A further aim is to investigate the level of QISS in cancer-related messages in the largest OHC in Slovenia and examine the differences among various types of discussion forums, namely, online consultation forums, online support group forums, and socializing forums. Methods: A multidimensional measurement instrument was developed, which included 20 items in a coding scheme for a content analysis of cancer-related messages. On a set of almost three million posts published between 2015 and 2019, a machine-learning algorithm was used to detect cancer-related discussions in the OHC. We then identified the messages providing informational social support, and through quantitative content analysis, three experts coded a random sample of 403 cancer-related messages for the QISS. Results: The results demonstrate a good level of interrater reliability and agreement for a QISS scale with six dimensions, each demonstrating good internal consistency. The results reveal large differences among the social support, socializing, and consultation forums, with the latter recording significantly higher quality in terms of accuracy (M = 4.48, P < .001), trust- worthiness (M = 4.65, P < .001), relevance (M = 3.59, P < .001), and justification (M = 3.81, P = .05) in messages providing informational social support regarding cancer-related issues. Conclusions: This study provides the research field with a valid tool to further investigate the factors and consequences of varying quality of information exchanged in supportive communication. From a practical perspective, OHCs should dedicate more resources and develop mechanisms for the professional moderation of health-related topics in socializing forums and thereby suppress the publication and dissemination of low-quality information among OHC users and visitors. Nasl. z nasl. zaslona. Opis vira z dne 23. 2. 2023. Bibliografija: str. 16-18. Abstract.

Supplemental material, sj-docx-1-dhj-10.1177_20552076231155681 for The quality of informational social support in online health communities: A content analysis of cancer-related discussions by Gregor Petrič, Marjan Cugmas, Rok Petrič and Sara Atanasova in Digital Health

Blockmodeling refers to a variety of statistical methods for reducing and simplifying large and complex networks. While methods for blockmodeling networks observed at one time point are well established, it is only recently that researchers have proposed several methods for analysing dynamic networks (i.e., networks observed at multiple time points). The considered approaches are based on k-means or stochastic blockmodeling, with different ways being used to model time dependency among time points. Their novelty means they have yet to be extensively compared and evaluated and the paper therefore aims to compare and evaluate them using Monte Carlo simulations. Different network characteristics are considered, including whether tie formation is random or governed by local network mechanisms. The results show the Dynamic Stochastic Blockmodel (Matias and Miele 2017) performs best if the blockmodel does not change otherwise, the Stochastic Blockmodel for Multipartite Networks (Bar-Hen et al. 2020) does.

Raziskovalci s področja družboslovja želijo pogosto razumeti odnos med družbenimi mikromehanizmi in družbenim makroizidom. V okviru analize omrežij so različni družbeni mikromehanizmi navadno operacionalizirani z lokalnimi omrežnimi mehanizmi, družbeni izidi na makroravni pa so operacionalizirani z različnimi globalnimi zgradbami omrežja (Stadtfeld, 2018). Tako je namen pričujoče disertacije proučiti odnos med lokalnimi omrežnimi mehanizmi in globalnimi omrežnimi zgradbami. Poleg nastanka izbranih vrst globalnih omrežnih zgradb je naslovljen tudi prehod iz ene v drugo globalno zgradbo omrežja. Različne vrste globalnih zgradb omrežij so opredeljene z različnimi vrstami bločnih modelov. Bločni model je definiran kot omrežje, v katerem so vozlišča skupine enakovrednih (glede na zgradbo povezav) vozlišč proučevanega omrežja, povezave pa so povezave med skupinami in znotraj skupin. Izraz »blok« se nanaša na matriko povezav, ki prikazuje povezave med vozlišči iz dveh različnih skupin ali med vozlišči znotraj ene skupine (Doreian in drugi, 2005). Tako kot različne globalne zgradbe omrežij je tudi družbene mehanizme mogoče opredeliti na različne načine. Skupna mnogim opredelitvam je trditev, da ima upoštevanje družbenih mehanizmov zelo pomembno vlogo pri pojasnjevanju družbenih pojavov (Hedström in Swedberg, 1998). Stadtfeld (2018) in Hedström in Swedberg (1998) so opredelili tri vrste mehanizmov: situacijske mehanizme (nanašajo se na vpliv globalne omrežne zgradbe na lastnosti posameznika, na primer na njegove želje, prepričanja in možnosti), vedenjske mehanizme (nanašajo se na to, kako lastnosti posameznika vplivajo na njegovo vedenje) in pretvorbene mehanizme (nanašajo se na to, kako vedenje posameznika vpliva na globalno omrežno zgradbo). V tej disertaciji so naslovljeni vedenjski ter pretvorbeni mehanizmi, ki jih je mogoče operacionalizirati prek množice pravil o vzpostavljanju, vzdrževanju in prekinjanju povezav. Disertacija je urejena v dveh delih. Zmožnost generiranja omrežij z izbrano vrsto bločnega modela, z upoštevanjem različnih vrst tirad, je naslovljena v prvem delu disertacije. To raziskovalno vprašanje je pomembno zlasti zato, ker še ni raziskave, ki bi sistematično naslovila povezavo med različnimi vrstami triad in različnimi vrstami bločnih modelov. Zmožnost generiranja omrežij z izbranimi vrstami bločnih modelov, brez upoštevanja lastnosti vozlišč, kaže, da lahko izbrane vrste bločnih modelov nastanejo kot rezultat izbranih lokalnih mehanizmov, kot so mehanizem popularnosti, mehanizem podobnosti stopenj, mehanizmi, povezani s tranzitivnostjo, in ostali. Za namene prvega raziskovalnega vprašanja so različne vrste triad razvrščene v množico dovoljenih ali v množico prepovedanih vrst triad. Takšna klasifikacija je narejena za vsako obravnavano vrsto bločnih modelov. Dana vrsta triade je dovoljena, če je njena frekvenca v omrežju z izbranim bločnim modelom brez napak večja od 0, sicer pa je prepovedana. Algoritem prestavljanja povezav (algoritem RL) ter algoritem MCMC sta bila uporabljena za generiranje omrežij z upoštevanje prepovedanih in/ali dovoljenih vrst triad. Z upoštevanjem različnih vrst triad je mogoče generirati večino analiziranih vrst bločnih modelov. To nakazuje, da se različne vrste bločnih modelov lahko pojavijo kot posledica lokalnih omrežnih mehanizmov, ki so neodvisni od lastnosti vozlišč. Drugi del disertacije obravnava lokalne omrežne mehanizme namesto lokalnih omrežnih zgradb, v okviru različnih vrst bločnih modelov. Medtem ko so lokalne omrežne zgradbe opredeljene z različnimi vrstami podomrežij, so lokalni omrežni mehanizmi procesi, ki vplivajo na konkretna dejanja vozlišč v omrežju, kot je to opisano v prejšnjih odstavkih. Različni lokalni omrežni mehanizmi so opredeljeni z različnimi omrežnimi statistikami, kar je upoštevano v algoritmih NEM. Ker obstaja veliko vrst bločnih modelov in lokalnih omrežnih mehanizmov, je za izbiro lokalnih omrežnih mehanizmov in pripadajočih vrst bločnih modelov dobro upoštevati izbrane družbene kontekste. V tej raziskavi sta upoštevana naslednja družbena konteksta: (i) prijateljstva in naklonjenosti med predšolskimi otroki ter (ii) pretok znanja med zaposlenimi v podjetju, ki temelji na znanju. Na osnovi navedenih družbenih kontekstov sta predlagani dve vrsti bločnih modelov: (simetričen ter asimetričen) središčno-koheziven bločni model ter hierarhično-koheziven bločni model z nekohezivno zadnjo skupino. V disertaciji je pokazana smiselnost obravnave obeh vrst bločnih modelov v okviru družbenih kontekstov, ki so povezani z vrtci ali s podjetji. Rezultati Monte Carlo simulacij kažejo, da lahko (simetričen in asimetričen) središčno-kohezivni bločni model nastane kot posledica mehanizmov vzajemnosti, popularnosti, podobnosti stopenj in mehanizmov, povezanih s tranzitivnostjo. To velja za vse obravnavane začetne zgradbe omrežij: prazno omrežje, omrežje s kohezivnim bločnim modelom ter omrežje z asimetričnim središčno-perifernim bločnim modelom. V disertaciji je pokazano (na osnovi že obstoječih podatkov, zbranih v ZDA), da se simetrična središčno-kohezivna vrsta bločnega modela pojavlja v omrežjih interakcij med predšolskimi otroki. To, da je mogoče omrežja s tako vrsto bločnega modela generirati z upoštevanjem navedenih lokalnih omrežnih mehanizmov, še ne pomeni, da so globalne zgradbe v empiričnih omrežjih nastale kot posledica analiziranih (v simulacijski študiji) lokalnih omrežnih mehanizmov. Ne glede na to, pojav take globalne zgradbe v omrežju prinaša nekatera pomembna vprašanja, povezana z (psihološkim) razvojem otrok, na katera je mogoče odgovoriti z upoštevanjem globalne zgradbe omrežja ter lastnosti otrok. Hierarhični bločni model z nekohezivno zadnjo skupino lahko nastane kot rezultat mehanizmov, povezanih s stroški, in mehanizmov, povezanih z vrednostjo. V tem primeru se stroški in vrednost navezujejo na dojemanje alterjevega znanja s strani ega (Nebus, 2006). Zmožnost generiranja globalnih omrežnih zgradb znotraj tega družbenega konteksta, z upoštevanjem lokalnih omrežnih mehanizmov, ki ne upoštevajo lastnosti vozlišč (z izjemo staža), nakazujejo na to, da je mogoče oblikovati take politike podjetja, ki spodbujajo nastanek želenega vzorca pretoka znanja (če tak želen vzorec pretoka znanja v podjetju obstaja). Osrednji prispevek disertacije je spoznanje, da lahko najbolj znane vrste bločnih modelov nastanejo kot rezultat zelo osnovnih lokalnih omrežnih mehanizmov, brez upoštevanja lastnosti vozlišč. Pri analizi razvoja bločnih modelov v empiričnih omrežjih je nujno upoštevati družbeni kontekst nastanka empiričnih omrežij ter vpliv družbenega konteksta na posameznikovo vedenje (Doreian in Conti, 2012). Social scientists often seek to understand the relationship between micro social mechanisms and macro social output. In the context of social networks, different micro social mechanisms are usually operationalized by local network mechanisms, while macro social outputs are operationalized by global network structures (Stadtfeld, 2018). Therefore, the aim of this dissertation is to study the relationship between local network mechanisms and global network structures. Not only is the emergence of the selected global network structures addressed, but so too is the transition from one global network structure to another. A blockmodel is used to define a global network structure. A blockmodel is defined as a network in which the nodes represent clusters of equivalent nodes (according to the structure of their links) from the studied network, while the links in a blockmodel represent the relationships between and within the clusters. The term “block” refers to a submatrix in an adjacency matrix that shows the relationships between nodes from two different clusters or between nodes from the same cluster (Doreian, Batagelj, & Ferligoj, 2005). Moreover, the social (network) mechanisms can be defined in different ways. Common to the various definitions is the claim that social mechanisms hold an important explanatory role (Hedström & Swedberg, 1998). Stadtfeld (2018) and Hedström & Swedberg (1998) defined three types of mechanisms: situational mechanisms (related to the global network structure’s impact on, e.g. the beliefs, desires and opportunities of an individual), action-formation mechanisms (associated with the impact of individuals’ beliefs, desires and opportunities on their actions/behaviour) and transformational mechanisms (related to the impact of individuals’ actions on the global network structure). In this study, the main focus is given to the last two types of local network mechanisms. The dissertation consists of two parts. The ability to generate networks with the selected blockmodel types, by considering only the triad types, is addressed in the first part. This research question is especially important because there is no known systematic study addressing a relationship between different triad types and blockmodels as the operationalization of global network structures. Whether the selected blockmodel types can be generated by considering only the triad types without any nodes’ attributes shows that these blockmodels can emerge as a consequence of local network mechanisms such as popularity, assortativity and others. To study the mentioned research question, different triad types are classified in the set of allowed and the set of forbidden triad types for each blockmodel type that is considered. A given triad type is called ‘allowed’ if its frequency in a given ideal blockmodel is higher than 0 otherwise, it is called ‘forbidden’. The proposed Relocating Links algorithm (RL algorithm) and the Markov Chain Monte Carlo algorithm (MCMC algorithm) are used to generate networks. In general, most studied blockmodels can be generated by only considering different triad types. This shows that some global network structures can emerge by virtue of the local network mechanisms that does not include the nodes’ attributes. The second part of the dissertation considers local network mechanisms, instead of local network structures, in the context of different blockmodel types. The local network mechanisms are processes that drive the specific actions of the nodes in the network, as described above. Different local network mechanisms are operationalized using different network statistics, which are considered by the nodes, when they obtain an opportunity to change the status of their links. This is done by different proposed algorithms from the NEM family. Given that there are many possible blockmodel types and possible local network mechanisms, the social context of the study is taken into account to select the most relevant blockmodel types and corresponding local network mechanisms. Two such social contexts considered in this dissertation are: (i) friendships and likings among pre-schoolers and (ii) the flow of knowledge among employees of an international, knowledge-based company. Based on these two social contexts, two blockmodel types are proposed: an (symmetric and asymmetric) core-cohesive blockmodel, and a hierarchical-cohesive blockmodel with last non-cohesive group. It is shown that the symmetric core-cohesive blockmodel type and the hierarchical-cohesive blockmodel with the last non-cohesive group are appropriate to be considered in the social context relating to a kindergarten and a company. The results of the Monte Carlo simulations show that the symmetric and asymmetric core-cohesive blockmodel types can emerge due to the mutuality, popularity, assortativity (of in-degree) and transitivity-related local network mechanisms when the initial global network structure is an empty network, a network with a cohesive blockmodel, or a network with an asymmetric core-periphery blockmodel. It was also shown (based on empirical data collected within a larger longitudinal study in the USA) that the symmetric core-cohesive blockmodel type appears in interactional networks among pre-schoolers. The fact this blockmodel type can be generated by the selected local network mechanisms does not imply that the global network structures of the empirical networks emerged due to the studied local network mechanisms. However, the appearance of this blockmodel type in the empirical data raises some very important developmental questions, which should be answered by considering the nodes’ attributes. A hierarchical-cohesive blockmodel with the last non-cohesive group can emerge as a result of so-called value-related mechanisms and cost-related mechanisms. Value and cost are defined through the ego's perception of the costs of obtaining the alter's knowledge and the value of the knowledge so obtained (Nebus, 2006). The ability to generate the global network structure, with local network mechanisms that do not consider the nodes’ attributes (except tenure), indicates that a company can develop policies that lead a knowledge flow towards the desired global structure (if it has one). The most important contribution of this dissertation is the observation that the most common blockmodel types can be generated by the basic local network mechanisms, without taking the attributes of the nodes into account. However, it is necessary to consider the social context and corresponding constraints on the nodes’ characteristics and their behaviour (Doreian & Conti, 2012) while analysing evolution of the global network in real networks.

Sodelovanje v znanosti igra pomembno vlogo tako v produkciji, kot v deseminaciji znanstvenega védenja. Čeprav so njegove meje težko določljive, je pogosto operacionalizirano skozi so-avtorstva znanstvenih bibliografskih enot, ki predstavljajo enega poglavitnih formalnih rezultatov znanstvenega sodelovanja. Iz osebnih bibliografij raziskovalcev je mogoče ustvariti tako imenovana so-avtorska omrežja, ki omogočajo preučevanje povezanosti nekaterih značilnosti raziskovalcev z vzorci vzpostavljanja so-avtorskih povezav, na ravni celotnega omrežja pa tudi ugotavljanje strukture tovrstnih omrežij. Z analizo so-avtorskih omrežij štirih znanstvenih disciplin v štirih petletnih obdobjih so Kronegger et al. (2011) z bločnim modeliranjem potrdili domnevo o strukturi tipa več-centrov—semi-periferija—periferija. Pričujoče delo analizo razširja na skoraj vse znanstvene discipline, kot jih opredeljuje Javna agencija za raziskovalno dejavnost Republike Slovenije (ARRS) in poleg strukture sodelovanja v znanosti naslovi še vprašanje stabilnosti znanstvenih sodelovanj skupin raziskovalcev glede na pripadnost vedi. Struktura so-avtorskih omrežij slovenskih raziskovalcev je preverjena z metodo neposrednega pred-določenega bločnega modeliranja, za merjenje stabilnosti dobljenih skupin raziskovalcev v času pa so vpeljane različne prilagoditve popravljenega Randovega indeksa. V kontekstu preučevanja so-avtorskih omrežij v dveh časovnih obdobjih, enote navadno prihajajo v omrežje (npr. mladi raziskovalci) ali ga zapuščajo (npr. upokojitev), kar pomeni, da je razvrščanje (bločno modeliranje) v prvem in v drugem časovnem obdobju izvedeno na dveh različnih množicah enot. Prilagojeni Randovi indeksi omogočajo primerjanje podobnosti dveh razvrstitev, ki sta izračunani na dveh množicah enot, kjer je ena množica enot podmnožica druge množice enot, združevanje in deljenje skupin v času pa različno vplivata na vrednost predstavljenih koeficientov. Predpostavljena struktura omrežja več-centrov—semi-periferija—periferija je značilna za vse analizirane discipline. Povprečna velikost dobljenih centrov je statistično značilno (p < 0,05) večja v prvem obdobju (5,6 raziskovalcev), v primerjavi z drugim obdobjem (4,4 raziskovalci). Glede na področje raziskovanja pa je povprečna velikost centrov statistično značilno (p < 0,05) višja v naravoslovno-tehniških disciplinah (4,6 raziskovalcev), kakor v družboslovno-humanističnih disciplinah (3,8 raziskovalcev). Stabilnost skupin raziskovalcev na ravni disciplin je relativno nizka in je prej posledica mnogih kratkoročnih sodelovanj, kakor prisotnosti deljenja raziskovalnih skupin. Na ravni ved je povprečna stabilnost disciplin statistično značilno večja v vedah Tehnika in Medicina v primerjavi s Humanistiko, medtem ko med združenimi vedami v skupini naravoslovno-tehniških ved in družboslovno-humanističnih ved ni razlike v povprečni stabilnosti dobljenih centrov. Collaboration in science plays an important role in the production as in the dissemination of a new scientific knowledge. Even there is hard to define the borders of scientific collaboration, the term is often operationalized through the co-authorship of scientific bibliographic units, which represents one of the most important results of a scientific collaboration. Based on the personal researchers’ bibliographies, the co-authorship networks can be constructed. These networks enable us to study the relationship between some researchers’ characteristics and the patterns of establishing new co-authorship ties. Furthermore, it allows us to study the structure of that kind of networks. Kronegger et al. (2011), who studied the co-authorship networks of four scientific disciplines in four five years periods, confirmed the hypothesis about the multi-core—semi-periphery—periphery structure. In the current work, the analysis is done on the level of almost all scientific disciplines, according to the Slovenian Research Agency (ARRS). Beside the structure of co-authorship networks, the current work also addresses the question of the stability of scientific collaboration teams across scientific fields. The structure of co-authorship networks of Slovenian researchers is examined using the pre-specified blockmodeling, while the stability of obtained clusters of researchers is measured with one of three proposed Modified Adjusted Rand Indices. In the context of co-authorship networks in two time periods, some researchers can enter or leave the network in the second time period. This implies that the classification (blockmodeling) is performed on two different sets of units for the first and for the second time period. The Modified Adjusted Rand Indices enable us to compare two clusterings, obtained on two different sets of units, where one set of units is a subset of another set of units. Moreover, the merging and splitting of clusters in time have a different effects on the value of proposed indices. The assumed network structure multi-core—semi-periphery—periphery exists in all analysed scientific disciplines. The average core size is statistically significantlly (p < 0.05) higher in the first time period (5.6 researchers) compared to the second time period (4.4 researchers). Depending on the field, the average core size is statistically significant (p < 0.05) higher in the fields of the natural and technical sciences (4.6 researchers) that in the fields of the Social sciences and Humanities (3.8 researchers). The stability of cores on the level of scientific disciplines is relatively low. Instability of cores is more the consequence of many short term collaborations rather than splitting of cores. On the level of scientific fields, the average stability of cores is statistically significant (p < 0.05) higher in the fields of the Engineering sciences and technologies and the Medical sciences in comparison to the Humanities, while on the level of merged scientific fields into the natural and technical sciences and social sciences and humanities, there is no difference in the average stability of obtained cores (the value of MARI1 is 0.21).

The purpose of this paper is to determine whether decision trees can be used to distinguish between those who are more and those who are less prone to depression and mental ill-being based on demographic and some other identifiable characteristics of individuals. Using decision trees we wanted to find segments which could be targeted in awareness-raising programs about depression. By analysing data from the European Social Survey in 2012 for Slovenia, we identified four groups at a greater risk of depression: young unemployed, unemployed passive, older people and women older than 52 years. Decision trees are also shown for finding target segments in the field of mental well-being. The results confirm the findings of previous studies. Namen prispevka je preveriti, ali lahko s pomočjo metode odločitvenih dreves na podlagi demografskih in nekaterih drugih določljivih značilnosti posameznika ločujemo med tistimi, ki so bolj, in tistimi, ki so manj nagnjeni k depresiji oziroma neugodnemu duševnemu počutju. Na ta način smo želeli poiskati ciljne segmente za ozaveščanje o depresiji in preventivnih programih. Študija je temeljila na analizi podatkov ESS za Slovenijo (leto 2012). Na podlagi analize smo identificirali štiri ciljne segmente, ki so v večji meri nagnjeni k depresiji: mladi brezposelni, brezposelni pasivni, starejši in ženske nad 52. letom starosti. Metoda odločitvenih dreves se je izkazala kot uporabna za iskanje ciljnih segmentov tudi na področju duševnega zdravja, rezultati pa potrjujejo izsledke drugih družboslovnih raziskav.

Introduction: Online counselling represents a new medium for finding health information. The aim of the research is to determine the importance of analysis of adolescents' issues in order to understand their problems, needs and desires. Methods: In 2012 the system for the classification of questions by the type of problem was introduced. In relation to the contents the questions were first sorted to the parent category then followed by the categorization according to the subject matter. The calculation comprised the portions, averages and quartiles, and in some cases even Cramer's V coefficients. The analysis covered the entire defined population (3,257 coded questions). Results: Most of the users are girls (76 %), the most representative group encompasses adolescents aged between 14 and 17 years (57 %). Most questions were grouped into the categories Sexuality and sexual health (24 %), Relationships (23 %) and Body (20 %). The length of posts increases with the age of the user (Cr's V = 0.18), but differs by the gender (a higher proportion of longer questions (Cr's V = 0.15) were posted by girls) and the themes (Cr's V = 0.31). Discussion and conclusion: The categorizing of questions is suitable for the identification and analysis of adolescents' problems, needs and desires. Regular categorisation of questions with analysis will serve as a useful research tool for youth work. Uvod: Spletne svetovalnice predstavljajo nov medij za iskanje informacij o zdravju. Cilj raziskave je ugotoviti pomen analiz vprašanj mladostnikov za razumevanje njihovih problemov, potreb in hrepenenj. Metode: Leta 2012 je bila uvedena katalogizacija vprašanj spletne svetovalnice glede na tipologijo problemov. Vprašanja so bila glede na vsebino sproti razvrščena v krovno kategorijo in nato pod več vsebinskih tem. Izračunani so bili deleži, povprečja ter kvartili, v nekaterih primerih Cramerjevih V koeficientov. Analiza je zajela celotno opredeljeno statistično populacijo (3.257 kodiranih vprašanj), obiskovalcev spletne svetovalnice, v obdobju med 1. januarjem 2012 in 31. decembrom 2012. Rezultati: Večino uporabnikov predstavljajo dekleta (76 %), najbolj reprezentativno skupino mladostnikov pa stari med 14 in 17 let (57 %). Največ vprašanj je bilo razvrščenih v kategorije spolnost in spolno zdravje (24 %), odnosi (23 %) in telo (20 %). Dolžina objav raste s starostjo uporabnika (Cr's V = 0,18), razlikuje pa se tudi glede na spol uporabnika (dekleta so objavila večji delež daljših vprašanj (Cr's V = 0,15)) in tematiko (Cr's V = 0,31). Diskusija in zaključek: Katalogizacija vprašanj je primerna za identifikacijo in analizo problemov, potreb in hrepenenj mladostnikov. Redna periodična klasifikacija vprašanj z analizami bo služila kot uporabno raziskovalno orodje za delo z mladostniki.

Avtizem je razvojna motnja, ki se začne izkazovati v obdobju malčkov (pri 14 mesecih), v povprečju pa je diagnosticirana pri 4 letih. Izražena je predvsem kot spremenjeno vedenje na področju socialne interackije, komunikacije in domišljije. Potreba po zgodnji diagnostiki avtizma ter potreba po časovni razbremenitvi diagnostičnega postopka sta glavna razloga za vedno pogostejšo implementacijo metod strojnega učenja na prvi stopnji diagnostičnega postopka, na kateri poteka identifikacija rizičnih enot. Cilja magistrskega dela sta identifikacija otrok z avtizmom in identifikacija najpomembnješih spremenljivk z uporabo napovednih klasifikacijskih modelov. V ta namen je bilo uporabljenih 6 različnih nadzorovanih metod strojnega učenja na 4 različnih starostnih skupinah. Vzorec zajema 48.050 otrok, starejših od 13 mesecev, od tega 60 % dečkov in 40 % deklic, pri čemer je diagnosticiranih s katerokoli diagnozo 20 % otrok. Pojavnost avtizma na vzorcu je 2 %. Uspešni smo bili pri identifikaciji avtizma v skupini otrok, starih od 37 do 48 mesecev, kjer je najuspešnejši model (metode naključnih gozdov) dosegel 72 % točnost, 59 % občutljivost za avtizem in 90 % specifičnost za avtizem. Model pri 10 % otrok brez avtizma napačno identificira znake avtizma. Pravilno identificira 59 % otrok z avtizmom, ostalim 41 % pa pripiše druge diagnoze. Za ocene metrik uspešnosti klasifikacije smo uporabili interno prečno preverjanje. Kot napomembnejši spremenljivki za klasifikacijo smo v skupini triletnikov identificirali spol in spremenljivki, ki merita otrokove slušne zaznave. Modeli so na splošno manj uspešni v mlajših starostnih skupinah, in sicer je v večini najmanj uspešen naiven Bayesov klasifikator. V najmlajših dveh skupinah smo bili najmanj uspešni pri identifikaciji otrok z avtizmom (9 % v najmlajši skupini in 18 % v skupini dvoletnikov). Uspešnost modelov v različnih starostnih skupinah se med seboj razlikuje, pri čemer ni videti jasnega trenda katerekoli metode, kar je lahko posledica različnih vprašalnikov in izraženosti avtizma v različnih starostnih skupinah. Povzamemo lahko, da navkljub veliki uporabnosti implementacije metod strojnega učenja pri iskanju rizičnih enot za avtizem nismo našli enovite rešitve, ki bi bila uspešna v vseh starostnih skupinah. Pri prepoznavanju diagnosticiranih otrok so modeli v vseh starostnih skupinah sicer zelo uspešni, a smo pri vseh kot največjo šibkost prepoznali slabo razločevanje avtizma od ostalih diagnoz. Autism is a developmental disorder with first signs in early childhood (at 14 months) and is typically diagnosed at the age of 4. It is primarily expressed as altered behavior in social interaction, communication and imagination. The need for early diagnosis of autism and the need for time-relieving the diagnostic process are the main reasons for the increasingly frequent implementation of machine learning methods in the first stage of the diagnostic process, which is the identification of children at risk . This study aims to identify children with autism using machine learning classification models and identify the most important variables for classification, using predictive classification models. For this purpose, 6 different supervised machine learning methods were used on 4 different age groups. The sample consisted of 48050 children over 13 months old, with 60 % boys and 40 % girls and 20 % diagnosed. The presence of autism in the sample was 2 %. The study was successful in identifying autism in a group of children aged 37-48 months where the most successful model (Random Forest method) achieved 72 % accuracy and 59 % sensitivity for autism and 90 % specificity for autism. The model misclassified 10 % of non-autistic cases and correctly identified 59 % of autistic cases and classified the remaining 41 % as having another diagnosis. Nested cross-validation was used for classification performance metrics estimation. In the group of three-year-olds, we identified gender and variables that measure children's auditory perception as the most important variables for classification. Models were generally less accurate in younger age groups, with the Naive Bayes classifier being the least accurate. The youngest two age groups however showed very low sensitivity for autism (9 % in the youngest group and 18 % in the two-year-old group). The evaluation of models in different age groups showed varying success and no clear trend for which method is the most successful across all groups, which could be a consequence of different screeners and expression of autism in different age groups. We can conclude that despite the growing use of machine learning methods in autism diagnosis, we were not successful with finding a definitive solution across different age groups. We can identify the main challenge: the models struggle to distinguish autism from other diagnoses and they successfully identify diagnosed cases but misclassify 40 % of autistic cases as having another diagnosis.