Maciej Grzenda

The evolutionary development of roughness prediction models

The vigorous expansion of wind energy power generation over the last decade has also entailed innovative improvements to surface roughness prediction models applied to high-torque milling operations. Artificial neural networks are the most widely used soft computing technique for the development of these prediction models. In this paper, we concentrate on the initial data transformation and its effect on the prediction of surface roughness in high-torque face milling operations. An extensive data set is generated from experiments performed under industrial conditions. The data set includes a very broad set of different parameters that influence surface roughness: cutting tool properties, machining parameters and cutting phenomena. Some of these parameters may potentially be related to the others or may only have a minor influence on the prediction model. Moreover, depending on the number of available records, the machine learning models may or may not be capable of modelling some of the underlying dependencies. Hence, the need to select an appropriate number of input signals and their matching prediction model configuration. A hybrid algorithm that combines a genetic algorithm with neural networks is proposed in this paper, in order to address the selection of relevant parameters and their appropriate transformation. The algorithm has been tested in a number of experiments performed under workshop conditions with data sets of different sizes to investigate the impact of available data on the selection of corresponding data transformation. Data set size has a direct influence on the accuracy of the prediction models for roughness modelling, but also on the use of individual parameters and transformed features. The results of the tests show significant improvements in the quality of prediction models constructed in this way. These improvements are evident when these models are compared with standard multilayer perceptrons trained with all the parameters and with data reduced through standard Principal Component Analysis practice.

On the prediction of floor identification credibility in RSS-based positioning techniques

The future of Location Based Services largely depends on the accuracy of positioning techniques. In the case of indoor positioning, frequently fingerprinting-based solutions are developed. A well known k Nearest Neighbours method is frequently used in this case. However, when the detection of a floor a mobile terminal is located at is an objective, only limited accuracy can be observed when the number of available signals is limited. The primary objective of this work is to analyse whether the credibility of floor estimates can be a priori assessed. A method assigning weights to individual GSM fingerprints and estimating their reliability in terms of floor estimation is proposed. The method is validated with an extensive radio map. It has been shown that both low and high accuracy floor estimates are correctly identified. Moreover, the objective criterion is proposed to assess individual weight functions from a proposed family of functions.

Prediction-oriented dimensionality reduction of industrial data sets

Soft computing techniques are frequently used to develop data-driven prediction models. When modelling of an industrial process is planned, experiments in a real production environment are frequently required to collect the data. As a consequence, in many cases the experimental data sets contain only limited number of valuable records acquired in expensive experiments. This is accompanied by a relatively high number of measured variables. Hence, the need for dimensionality reduction of many industrial data sets. The primary objective of this study is to experimentally assess one of the most popular approaches based on the use of principal component analysis and multilayer perceptrons. The way the reduced dimension could be determined is investigated. A method aiming to control the dimensionality reduction process in view of model prediction error is evaluated. The proposed method is tested on two industrial data sets. The prediction improvement arising from the proposed technique is discussed.

Heat Consumption Prediction with Multiple Hybrid Models

Load forecasting plays an important role in modern utilities. However, further improvements can be expected by predicting the load at a consumer level. The latter approach has become available with the advent of low-cost monitoring and transmission systems. Still, due to the limited number of monitored clients, the way groups of consumers should be identified and whether their data is sufficient for high quality prediction models remains an open issue. The work summarises the results of building prediction models for different consumer groups of a district heating system. The way self-organising maps, multilayer perceptrons and simple prediction strategies can be applied to identify groups of consumers and build their prediction models has been proposed. The hypothesis that a billing database enables group identification has been verified. Significant improvements in prediction accuracy have been observed.

Interpreting tree-based prediction models and their data in machining processes

Machine-learning techniques frequently predict the results of machining processes, based on pre-determined cutting tool settings. By doing so, key parameters of a machined product can be predicted before production begins. Nevertheless, a prediction model cannot capture all the features of interest under real-life industrial conditions. Moreover, careful assessment of prediction credibility is necessary for accurate calibration; aspects that should be addressed through appropriate modeling and visualization techniques. A machine process test problem is proposed to analyze data-visualization techniques, in which a real data set is analyzed that describes deep-drilling under different cutting and cooling conditions. The main objective is the efficient fusion of visualization techniques with the knowledge of industrial engineers. Common modeling and visualization techniques were first surveyed, to contrast standard practice with our novel approach. A hybrid technique combining conditional inference trees with dimensionality reduction was then examined. The results show that a process engineer will be able to estimate overall model accuracy and to verify the extent to which accuracy depends on industrial process settings and the statistical significance of model predictions. Moreover, evaluation of the data set in terms of its sufficiency for modeling purposes will help assess the credibility of these decisions.

Towards the reduction of data used for the classification of network flows

The ever growing volume of network traffic results in the need for even more efficient data processing in Intrusion Detection Systems. In particular, the raw network data has to be transformed and largely reduced to be processed by data mining models. The primary objective of this work is to control the dimensionality reduction (DR) of network flow records in view of the accuracy of misuse detection. A real data set, containing flow records with potential spam messages, is used to perform the tests of the proposed method. The algorithm proposed in this study is applied to investigate the merits of hybrid models composed of dimensionality reduction, neural networks, and decision trees. The benefits of dimensionality reduction and the impact of the process on the overall spam detection rates and false positive rates are investigated. The advantages of the proposed technique over standard a priori selection of reduced dimension are discussed.

Handling incomplete data using evolution of imputation methods

In this paper new approach to treat incomplete data has been proposed. It has been based on the evolution of imputation strategies built using both non-parametric and parametric imputation methods. Genetic algorithms and multilayer perceptrons have been applied to develop a framework for constructing the imputation strategies addressing multiple incomplete attributes. Furthermore we evaluate imputation methods in the context of not only the data they are applied to, but also the model using the data. The accuracy of classification on data sets completed using obtained imputation strategies has been described. The results outperform the corresponding results calculated for the same data sets completed using standard strategies.

Requirements and Solutions for Web-Based Expert System

The advent of the Internet has strongly influenced modern software systems. Existing intranet solutions are being gradually replaced with www services available everywhere and at any time. The availability of the wide area network has resulted in unprecedented opportunities of remote and distributed cooperation of large groups of people and organizations.

The Adoption of Open Data and Open API Telecommunication Functions by Software Developers

The primary objective of the work is the preliminary investigation of the adoption of Open Data and Open API telecommunication functions by software developers. The analysis is based on the statistical data collected during developer contests. Based on Open API Hackathon and Business Intelligence API Hackathon contests, the interest of software developers in telecommunication operator functions and Open Data exposed in Open APIs form is assessed. Conclusions on the categories of open city data attracting the attention of software developers are formulated.

Dynamic Transfer Patterns for Fast Multi-modal Route Planning

Route planning makes direct use of geographic data and provides beneficial recommendations to the public. In real-world the schedule of transit vehicles is dynamic and delays in the schedules occur. Incorporation of these dynamic schedule changes in multi-modal route computation is difficult and requires a lot of computational resources. Our approach extends the state-of-the-art for static transit schedules, Transfer Patterns, for the dynamic case. Therefore, we amend the patterns by additional edges that cover the dynamics. Our approach is implemented in the open-source routing framework OpenTripPlanner and compared to existing methods in the city of Warsaw. Our results are an order of magnitude faster then existing methods.