Browsing by Author "Ala-Fossi, Mikko"

Now showing 1 - 2 of 2

Adaptation layer software for an ATM switch
(2002) Ala-Fossi, Mikko
Helsinki University of Technology | Master's thesis
Asynchronous Transfer Mode (ATM) on nykyään laajalti käytetty verkkotekniikka. ATM-kytkimen, jonka tehtävänä on kytkeä muut verkkoelementit ATM-verkossa, täytyy senkin tarjota oma ATM-verkkoliittymänsä. ATM-sovituskerros (AAL) yhdistää ylempien kerroksien protokollat ja sovellukset tähän verkkoliittymään, ja hoitaa näiden tiedonsiirtoyksiköiden siirron käyttäen ATM-kerroksen tiedonsiirtoyksiköitä soluja. Tässä diplomityössä selvitetään ATM-sovituskerroksen ohjelmistoarkkitehtuuri ja toteutus ATM-kytkimessä, esitellään siihen liittyvät laite- ja ohjelmistoseikat sekä ATM-tekniikan yleiset periaatteet. Lisäksi selvitetään sovituskerroksen segmentointi- ja uudelleenkoontitason (SAR) toteutus Conexant:n RS8234-piirille ja konvergenssikerroksen yleisosan (CPCS) toteutus ATM-kytkimessä.
Identifying Interesting Episode Patterns in User Interaction Log Data
(2023-01-23) Rinta-Paavola, Juho
Perustieteiden korkeakoulu | Master's thesis
Introduction: Axel Encounter is a patient flow management system that helps patients and healthcare professionals to be at the right place at the right time, while reducing burden of routine tasks and making the experience smooth. In this thesis project, we developed an episode mining framework and used it to analyze how healthcare professionals interact with the system. Methods: We used the framework to discover episodes and episode rules in user interaction log data from six healthcare organizations. The framework was based on an enhanced version of the Winepi algorithm. We used goodness measures that rank episode rules based on statistical dependence, addressing shortcomings of the previously used measure. We pruned superfluous rules that provided no information not provided by a simpler rule. We proved that some episodes can only be involved in superfluous rules, giving a novel way to prune the search space. Results: We discovered episode rules that provided practically relevant information. All goodness measures should be considered, noting their advantages and disadvantages. Pruning reduced the number of discovered rules vastly and provided a large performance boost, which made the results easier to interpret and the experiments feasible. Conclusions: Episode mining is a viable method for analyzing log data. Suitable goodness measures and pruning are important for discovering rules that are easy to interpret. We propose further improvements, such as pruning more aggressively, and extending goodness measures and pruning methods to work with more general definitions of episodes and episode rules. We discuss issues of frequency-based episode mining.