Araştırma Çıktıları | WoS | Scopus | TR-Dizin | PubMed

Permanent URI for this communityhttps://hdl.handle.net/20.500.14719/1741

Browse

Filters

2020 - 20243

Settings

Start date: 2020End date: 2026Type: search.filters.itemtype.Proceedings PaperAuthor: search.filters.author.Soyak, Ece GelalSubject: search.filters.subject.Wi-Fi

Search Results

Now showing 1 - 3 of 3
  • No Thumbnail Available
    PublicationMetadata only
    Advancing WebRTC QoE Assessment with Machine Learning in Real-World Wi-Fi Scenarios
    (IEEE, 2024) Argin, Berke; Demir, Mehmet Ozgun; Salik, Elif Dilek; Onalan, Aysun Gurur; Batum, Oyku Han; Soyak, Ece Gelal; Bahcesehir University
    Video conferencing applications play a key role in enabling use cases like remote working, education, and potentially the metaverse. From the perspective of Internet service providers, predicting the end user's Quality of Experience (QoE) in such applications is critical in allocating the right resources to ensure consistently high QoE. This work addresses the estimation of user QoE from link-layer performance metrics such as transferred packets, queue size, signal strength, and channel occupancy for WebRTC-supported applications. Our study entails collecting a data set capturing various Wi-Fi scenarios in practical environments and training machine learning models on this data to estimate the perceived QoE. Our findings demonstrate improvement in prediction accuracy compared to earlier models and QoE representations, furthermore, we also investigate the explainability of the models with the help of SHAP values.
  • No Thumbnail Available
    PublicationMetadata only
    Faster Wi-Fi Fingerprinting Using Feature Selection
    (IEEE, 2020) Aydin, Hurkan M.; Ali, Muhammad Ammar; Soyak, Ece Gelal; Bahcesehir University; Bahcesehir University
    Wi-Fi fingerprinting has been widely used for indoor positioning, as Wi-Fi technology is easily deployed and supported. In fingerprinting, a database is created using the received signal strength indicator (RSSI) values in the area of interest, position prediction is performed by finding the best match for a measured RSSI among the values in the database. As location positioning gains importance for continuous interactive (CI) applications in large indoor spaces such as malls and airports, the fingerprinting databases become larger, making it computationally more difficult to position targets in real-time. On the other hand, CI applications such as Augmented Reality (AR) require low-latency positioning for a good user experience. In this work, we propose to use feature selection methods along with the K-nearest neighbors (KNN) classification and regression algorithms in order to create a simple and swift location positioning system. Our evaluation of various feature selection methods shows that computation times for positioning can be reduced by 75% using feature selection.
  • No Thumbnail Available
    PublicationMetadata only
    The Analysis of Feature Selection with Machine Learning for Indoor Positioning
    (IEEE, 2021) Aydin, Hurkan M.; Ali, Muhammad Ammar; Soyak, Ece Gelal; Bahcesehir University; Bahcesehir University
    Indoor positioning is useful in various venues including warehouses, convention centers, malls, airports, nursing homes. In these scenarios, reducing the complexity of location estimation both improves responsiveness and helps to elongate battery life of the mobile device. In this work, we carry out a detailed analysis of the impact of Principal Component Analysis (PCA) on the computational complexity and accuracy with different machine learning algorithms on a large data set containing 520 APs. We compare the algorithms' training and testing times, as well as their accuracies in the presence and absence of PCA. Our results show that (i) PCA significantly reduces both the training and testing times for classification and regression using k-nearest neighbor (kNN) and support vector machine (SVM) algorithms while preserving if not improving accuracy, (ii) PCA slightly improves the training/testing times for regression using multi-layer perceptron (MLP), (iii) random forest (RF) does not perform well with PCA.