103,434 research outputs found
Primerjava algoritmov za napovedno analitiko podatkov
No full textThe master’s degree thesis is composed of theoretical and practical parts. The theoretical part describes the basics of predictive data analytics and machine learning algorithms for classification such as Logistic Regression, Decision Tree, Random Forest, SVM, and KNN. We also describe different evaluation metrics such as Recall, Precision, Accuracy, F1 Score, Cohen’s Kappa, Hamming Loss, and Jaccard Index that are used to measure the performance of these algorithms. Additionally, we record the time taken for the training and prediction processes to provide insights into algorithm scalability.
The key part master’s thesis is the practical part that compares these algorithms with a self-implemented tool that shows results for different evaluation metrics on seven datasets. First, we describe the implementation of an application for testing where we measure evaluation metrics scores. We tested these algorithms on all seven datasets using Python libraries such as scikit-learn. Finally, wMagistrsko delo je sestavljeno iz teoretičnega in praktičnega dela, ki skupaj omogočata celovit pregled nad področjem napovedne analitike podatkov in algoritmov strojnega učenja za klasifikacijo. V teoretičnem delu se poglobljeno ukvarjamo z osnovami napovedne analitike, kjer podrobno obravnavamo glavne algoritme strojnega učenja, ki se uporabljajo za klasifikacijo podatkov. Med temi algoritmi so logistična regresija, odločitveno drevo, naključni gozd, podporni vektorski stroj (SVM) in k-najbližjih sosedov (KNN). Vsakega od teh algoritmov podrobno predstavimo z vidika njihovega delovanja, prednosti in pomanjkljivosti ter primerov uporabe, kje so najučinkovitejši. Posebna pozornost je namenjena razumevanju, kako ti algoritmi obdelujejo podatke ter kako prilagoditev njihovih parametrov vpliva na končne rezultate, kar je ključno za optimalno uporabo v različnih kontekstih in situacijah. Hkrati pa se poglobimo v teoretične osnove delovanja teh algoritmov, kar omogoča boljše razumevanje njihove praktične uporabe v različnih scenarijih.
V teoretičnem delu prav tako podrobno obravnavamo različne metrike ocenjevanja, ki so ključne za merjenje uspešnosti algoritmov strojnega učenja. Te metrike vključujejo priklic, natančnost, točnost, F1 rezultat, Cohenov Kappa, Hammingova izguba in Jaccardov indeks. Vsako od teh metrik natančno predstavimo in pojasnimo njihovo matematično ozadje ter njihov vpliv na oceno delovanja algoritmov v različnih situacijah, še posebej v primerih, ko so podatki neuravnoteženi ali ko so kriteriji za uspešnost drugačni od običajnih. Posebno pozornost namenjamo tudi časovni učinkovitosti algoritmov, saj čas, potreben za učenje in napovedovanje, ponuja pomemben vpogled v njihovo razširljivost in primernost za uporabo na velikih podatkovnih nizih. Prav tako smo izpostavili pomembnost prilagoditve teh metrik glede na specifične zahteve različnih podatkovnih nizov in analitičnih ciljev, kar je ključno za pravilno interpretacijo rezultatov.
Praktični del magistrske naloge je osredotočen na implementacijo in primerjavo omenjenih algoritmov v realnem okolju. Razvili smo orodje, ki omogoča samostojno izvajanje testov in prikaz rezultatov za različne metrike ocenjevanja na sedmih različnih podatkovnih nizih. V tem delu natančno opisujemo postopek razvoja in implementacije te aplikacije za testiranje, pri čemer smo uporabili več Python knjižnic, med njimi scikit-learn, ki je osrednja knjižnica za strojno učenje v Pythonu. Algoritme smo testirali na vseh sedmih podatkovnih nizih, pri čemer smo posebno pozornost namenili časovni učinkovitosti in natančnosti rezultatov. Ta dva dejavnika neposredno vplivata na razširljivost in uporabnost teh metod v realnih aplikacijah, kar je ključno za nadaljnji razvoj in uporabo tehnik strojnega učenja. Poleg tega smo v tem delu analizirali, kako razlike v podatkovnih nizih vplivajo na učinkovitost algoritmov, kar je pomembno za njihovo izbiro v specifičnih scenarijih uporabe.
V zaključnem delu magistrske naloge smo izvedli poglobljeno analizo pridobljenih rezultatov, kjer smo primerjali učinkovitost posameznih algoritmov glede na različne metrike ocenjevanja. Na podlagi te analize smo oblikovali zaključke, ki nudijo poglobljen vpogled v prednosti in omejitve uporabljenih algoritmov. Ugotovili smo, da so logistična regresija, odločitveno drevo, in naključni gozd na večini testiranih podatkovnih nizov izkazali izjemno zmogljivost. Nasprotno so algoritmi, kot sta SVM in KNN, v določenih primerih dosegli nekoliko nižje ocene, kar kaže na potrebo po previdni izbiri algoritma glede na specifične značilnosti podatkov. Poleg tega smo podali priporočila za njihovo optimalno uporabo v prihodnjih raziskavah in realnih aplikacijah na področju napovedne analitike in strojnega učenja, pri čemer smo upoštevali tudi pomembne vidike, kot so robustnost, prilagodljivost, časovna učinkovitost ter praktična uporabnost teh algoritmov v različnih industrijskih sektorjih
Comparison of databases regarding saving JSON documents
No full textDiplomsko delo je sestavljeno iz teoretičnega in praktičnega dela. Najprej so opisane osnove relacijskih in nerelacijskih podatkovnih baz, nato pa njihovi najbolj znani predstavniki. Nato sledi razlaga formata za izmenjavo podatkov JSON in dela z njim v podatkovnih bazah (ustvarjanje, branje, posodabljanje, brisanje podatkov).
Ključni del diplomske naloge je praktično delo, kjer smo merili in analizirali podatkovne baze pri shranjevanju dokumentov JSON. Najprej opišemo implementacijo aplikacije za samodejno testiranje, kjer merimo čas in porabo pomnilnika. Testiranje je bilo izvedeno nad manjšimi in večjimi dokumenti. Testirali smo podatkovne baze MySQL, PostgreSQL in MongoDB. Na koncu analiziramo dobljene rezultate in podamo zaključne ugotovitve.The diploma thesis consists of theoretical and practical work. The basics of relational and non-relational databases are described first, followed by their most well-known representatives. This is followed by an explanation of the JSON data exchange format and working with it in databases (creating, reading, updating, deleting data).
The key part of the thesis is practical work, where we measured and analyzed databases when storing JSON documents. We first describe the implementation of an automated testing application which measures time and memory consumption. Testing was performed on smaller and larger documents. We tested MySQL, PostgreSQL and MongoDB databases. Finally, we analyze the obtained results and give concluding remarks
Direct analysis of the genes encoding G proteins G alpha T2, G alpha o, G alpha Z in ADHD
No full textWe have followed up the extensive replicated evidence that the dopamine DRD4 receptor is involved in the aetiology of ADHD by undertaking direct analysis of genes encoding other proteins in this effector system. We prioritised the genes encoding G protein subunits GT2, Go, GZ as these have been shown to transduce the effects of ligand binding at DRD4. We screened the exons of all three genes for sequence variation in 28 unrelated subjects with ADHD and identified 13 novel polymorphisms. All were tested for possible association with ADHD using a combination of pooled and individual genotyping. The results of our study do not suggest that polymorphisms in these genes contribute to susceptibility to ADH
REM-sleep changes in children with attention-deficit/hyperactivity disorder: Methodologic and neurobiologic considerations
No full text
Automated genotyping of single-nucleotide polymorphisms by extension of fluorescently labelled primers: analysis of individual and pooled DNA samples
No full textKirov, G. Stephens, M. Williams, N.M. ODonovan, M.C. Owen, M.J
Family-based association studies of candidate genes in bipolar disorder
No full textKirov G, Jones I, McCandless F, Craddock N, Owen M
The MAthSAT Solver. A progress report
No full textMany problems of practical relevance are conveniently expressed as boolean combinations of propositional variables and mathematical constraints. The development of decision procedures able to check the satisfiability of such formulas is therefore being devoted an increasing interest.
The MathSat family of deciders is based on the extension of a DPLL propositional satisfiability procedure, used as an assignment enumerator. MathSat pioneers a lazy and layered approach, where propositional reasoning is tightly integrated with solvers of increasing expressive power (e.g. to reason about equality and linear arithmetic) in such a way that ``more expensive'' layers are called less frequently.
In this paper, we show the advances in the development of MathSat. We discuss the implications related to the use of Minisat, a new-generation propositional SAT solver; the role of an incremental mathematical reasoner; the role of static learning; and the extension to integer variables. We show that the new version of MathSat is significantly more efficient than the previous on
A Genomic Analysis of the Bacillus Bacteriophage Kirovirus kirovense Kirov and Its Ability to Preserve Milk
No full textBacteriophages are widely recognized as alternatives to traditional antibiotics commonly used in the treatment of bacterial infection diseases and in the food industry, as phages offer a potential solution in combating multidrug-resistant bacterial pathogens. In this study, we describe a novel bacteriophage, Kirovirus kirovense Kirov, which infects members of the Bacillus cereus group. Kirovirus kirovense Kirov is a broad-host-range phage belonging to the Caudoviricetes class. Its chromosome is a linear 165,667 bp double-stranded DNA molecule that contains two short, direct terminal repeats, each 284 bp long. According to bioinformatics predictions, the genomic DNA contains 275 protein-coding genes and 5 tRNA genes. A comparative genomic analysis suggests that Kirovirus kirovense Kirov is a novel species within the Kirovirus genus, belonging to the Andregratiavirinae subfamily. Kirovirus kirovense Kirov demonstrates the ability to preserve and decontaminate B. cereus from cow milk when present in milk at a concentration of 104 PFU/mL. After 4 h of incubation with the phage, the bacterial titer drops from 105 to less than 102 CFU/mL
A System for Assessing the Knowledge and Skills of Students in Computer Programming
No full textThis article presents the author’s experience in
teaching programming to university students in first and second year. The system for assessing the knowledge and skills of students is an essential part of teaching. It aims not only to assess students, but also to help improving their knowledge. The importance and difficulty of Computer Programming requires specific and unconventional approach to this activity. The article discusses all three elements of ongoing assessment (tests for knowledge, homeworks, and practical programming skills), as well as the rules for final exams and final assessment
- …
