|
|
Machine Learning Optimization of KPI Prediction
Haris, Daniel ; Burget, Radek (referee) ; Bartík, Vladimír (advisor)
This thesis aims to optimize the machine learning algorithms for predicting KPI metrics for an organization. The organization is predicting whether projects meet planned deadlines of the last phase of development process using machine learning. The work focuses on the analysis of prediction models and sets the goal of selecting new candidate models for the prediction system. We have implemented a system that automatically selects the best feature variables for learning. Trained models were evaluated by several performance metrics and the best candidates were chosen for the prediction. Candidate models achieved higher accuracy, which means, that the prediction system provides more reliable responses. We suggested other improvements that could increase the accuracy of the forecast.
|
|
|
Knowledge Discovery from Web Logs
Vlk, Vladimír ; Očenášek, Pavel (referee) ; Bartík, Vladimír (advisor)
This master's thesis deals with creating of an application, goal of which is to perform data preprocessing of web logs and finding association rules in them. The first part deals with the concept of Web mining. The second part is devoted to Web usage mining and notions related to it. The third part deals with design of the application. The forth section is devoted to describing the implementation of the application. The last section deals with experimentation with the application and results interpretation.
|
|
|
Image processing with neural networks
Gróf, Zoltán ; Pohl, Jan (referee) ; Jirsík, Václav (advisor)
This bachelor’s thesis centralizes on the possible uses of neural networks in the field of computer vision. This work contains basic theoretic knowledge of the field of neural networks and image processing. It discusses how successfully can neural networks be applied through the separate steps of image processing, what kind of neural networks are suitable for these steps, and what are the problems that might appear with their use. The work discusses the fields of classification and image understanding in a more detailed level. It’s shown how the use of neural networks can be appropriate in these applications. An own program was created as part of this work to demonstrate the classification capabilities of neural networks. It’s shown a neural network is created and trained for the recognition of handwritten numbers. The trained neural network was subject to different tests, through which the conclusion was reached, that it works with a high success rate, but is sensitive to changes in the input objects: change of size and location. A number of possible solutions were designed for this problem.
|
|
|
Basic fingerprint liveness detection
Horák, Tomáš ; Smital, Lukáš (referee) ; Kašpar, Jakub (advisor)
This work focuses on recognizing fingerprints liveness based purely on software-methods evaluating symptoms from just one fingerprint image. At first in this work was described the issue of biometry as such, comparing the advantages and disadvantages of such systems. Next part deal with more detailed process of fingerprint biometry including papillary lines and overall fingerprints as such. In the next phase, the problems and utilization of both software and hardware methods are discussed, including principles of individual approaches. This part is followed by a selection of used fingertip symptoms. This is followed by the practical part and the LivDet 2011 database, which was used for finger recognition. In the practical part is also described the used neural network capturing minor differences in fingerprints according to 13 symptoms.
|
|
|
Web Mining - Clustering
Rychnovský, Martin ; Burget, Radek (referee) ; Bartík, Vladimír (advisor)
This work presents the topic of data mining on the web. It is focused on clustering. The aim of this project was to study the field of clustering and to implement clustering through the k-means algorithm. Then, the algorithm was tested on a dataset of text documents and on data extracted from web. This clustering method was implemented by means of Java technologies.
|
|
| |
|
|
Segmentation in microscopic images
Vlk, Jaroslav ; Odstrčilík, Jan (referee) ; Kolář, Radim (advisor)
Work studying the properties of microscopic images and then the following applies in the segmentation of the image. Work is trying to use a simpler methods image processing mainly. At the same time, deals with methods for preprocessing image. In the choice of methods and the use of emphasis on speed and simplicity of calculation.
|
|
|
Manifestation of physiological noise in fMRI data
Skoupý, Radim ; Mareček, Radek (referee) ; Lamoš, Martin (advisor)
In this bachelor thesis I deal with the manifestations of physiological noise in fMRI data. The work includes the basic theory of BOLD signal - his character, dealing with the processing of measured data, the possibilities of their treatment, filtration and formation of the resulting statistical parametric maps. The key part is studing possibilities of physiological noise filtering method Retroicor that models physiological noise based on sine and cosine basis functions.
|
|
|
Textual Data Clustering Methods
Miloš, Roman ; Burgetová, Ivana (referee) ; Bartík, Vladimír (advisor)
Clustering of text data is one of tasks of text mining. It divides documents into the different categories that are based on their similarities. These categories help to easily search in the documents. This thesis describes the current methods that are used for the text document clustering. From these methods we chose Simultaneous keyword identification and clustering of text documents (SKWIC). It should achieve better results than the standard clustering algorithms such as k-means. There is designed and implemented an application for this algorithm. In the end, we compare SKWIC with a k-means algorithm.
|
|
|
Automatic acquisition of values from measurement devices without communication interface
Dohnálek, Martin ; Čala, Martin (referee) ; Kunz, Jan (advisor)
This bachelor thesis deals with the matter of optical character recognition from displays of measurement devices without communication interface. This would allow carrying out automated experiments using cheaper or older gear, which is not endowed with means for direct connection to a computer. Input image necessary for the character recognition is acquired using a camera pointed at a display of the device. The recognition is afterwards performed on periodically captured image based on an already existing dataset for particular apparatus. The output of the algorithm is a file containing recognized values, units, and timestamps of the recognition. The tool for creating datasets was designed as well. The achieved speed of recognition (as fast as 34 ms per iteration) during practical testing confirmed the sufficient optimalization of OCR algorithm. On the other hand, the determined hit rate of recognition abiding specified conditions was nearly 100 %. Lastly, the resistance to misalignment of display and sensor plane was monitored. The OCR algorithm is resilient to horizontal tilt up to +/- 5° and vertical tilt up to +/- 20°.
|
guest ::
