Numerous real-world data contain missing values, while in contrast, most Machine Learning (ML) algorithms assume complete datasets. For this reason, several imputation algorithms have been proposed to predict and fill in the missing values. Given the advances in predictive modeling algorithms tuned in an Automated Machine Learning context (AutoML) setting, a question that naturally arises is to what extent sophisticated imputation algorithms (e.g., Neural Network based) are really needed, or we can obtain a descent performance using simple methods like Mean/Mode (MM). In this article, we experimentally compare six state-of-the-art representatives of different imputation algorithmic families from an AutoML predictive modeling perspective, including a feature selection step and combined algorithm and hyper-parameter selection. We used a commercial AutoML tool for our experiments, in which we included the selected imputation methods. Experiments ran on 25 binary classification real-world incomplete datasets with missing values and 10 binary classification complete datasets in which synthetic missing values are introduced according to different missingness mechanisms, at varying missing frequencies. The main conclusion drawn from our experiments is that the best method on average is the Denoise AutoEncoder on real-world datasets and the MissForest in simulated datasets, followed closely by MM. In addition, binary indicator variables encoding missingness patterns actually improve predictive performance, on average. Last, although there are cases where Neural-Network-based imputation significantly improves predictive performance, this comes at a great computational cost and requires measuring all feature values to impute new samples.

MOTIVATION: Genome-wide association studies (GWAS) present several computational and statistical challenges for their data analysis, including knowledge discovery, interpretability, and translation to clinical practice. RESULTS: We develop, apply, and comparatively evaluate an automated machine learning (AutoML) approach, customized for genomic data that delivers reliable predictive and diagnostic models, the set of genetic variants that are important for predictions (called a biosignature), and an estimate of the out-of-sample predictive power. This AutoML approach discovers variants with higher predictive performance compared to standard GWAS methods, computes an individual risk prediction score, generalizes to new, unseen data, is shown to better differentiate causal variants from other highly correlated variants, and enhances knowledge discovery and interpretability by reporting multiple equivalent biosignatures. AVAILABILITY AND IMPLEMENTATION: Code for this study is available at: https://github.com/mensxmachina/autoML-GWAS. JADBio offers a free version at: https://jadbio.com/sign-up/. SNP data can be downloaded from the EGA repository (https://ega-archive.org/). PRS data are found at: https://www.aicrowd.com/challenges/opensnp-height-prediction. Simulation data to study population structure can be found at: https://easygwas.ethz.ch/data/public/dataset/view/1/.

Electrocardiogram (ECG) signal constitutes a valuable technique that provides considerable information towards the early diagnosis of several cardiovascular diseases, especially regarding the detection of abnormal heart rate, namely arrhythmias. In this paper, innovative methodologies that allow for the efficient classification of cardiac rhythm are presented. The proposed methods are based on ECG signal analysis, extraction of significant features, as well as classification algorithms. Several clinical, time- and frequency-domain features are either calculated, or automatically extracted by means of a Convolutional Neural Network, while traditional machine learning algorithms, such as k-Nearest Neighbors and Random Forests are employed in order to classify the ECG signals among 7 different cases of abnormal and normal heart rate. The learning methods are carried out within the JADBio software tool, that also performs feature selection prior to classification. The experimental results demonstrate high performance of the deployed methods in terms of relevant statistical metrics, while they yielded an average validation Area Under the Curve (AUC) of 99.9%.

Depression is one of the most prominent mental disorders, with an increasing rate that makes it the fourth cause of disability worldwide. The field of automated depression assessment has emerged to aid clinicians in the form of a decision support system. Such a system could assist as a pre-screening tool, or even for monitoring high risk populations. Related work most commonly involves multimodal approaches, typically combining audio and visual signals to identify depression presence and/or severity. The current study explores categorical assessment of depression using audio features alone. Specifically, since depression-related vocal characteristics impact the glottal source signal, we examine Phase Distortion Deviation which has previously been applied to the recognition of voice qualities such as hoarseness, breathiness and creakiness, some of which are thought to be features of depressed speech. The proposed method uses as features DCT-coefficients of the Phase Distortion Deviation for each frequency band. An automated machine learning tool, Just Add Data, is used to classify speech samples. The method is evaluated on a benchmark dataset (AVEC2014), in two conditions: read-speech and spontaneous-speech. Our findings indicate that Phase Distortion Deviation is a promising audio-only feature for automated detection and assessment of depressed speech.

Assessing the impact of different network conditions on user experience is important for improving the telecommunication services. We have developed a modular framework that includes monitoring and data collection tools and algorithms for user-centric analysis and prediction of the QoE in video streaming. The MLQoE employs several machine learning (ML) algorithms and tunes their hyper-parameters. It dynamically selects the ML algorithm that exhibits the best performance and its parameters automatically based on the input (e.g., network and systems metrics). We applied the MLQoE for predicting the QoE of the video streaming service in the context of two field studies, one performed in the production environment of a large telecom operator and the other at our Institute. The analysis indicated the parameters with the dominant impact on the perceived QoE and revealed that the QoE vary across users. This motivates the use of customized adaptation mechanisms in video streaming under network performance degradation. The MLQoE results in fairly accurate predictions e.g., a median error in predicting the QoE of 0.0991 and 0.5517 in the first (second) field study, respectively, on the MOS scale.

The impact of the network performance on the quality of experience (QoE) for various services is not well-understood. Assessing the impact of different network and channel conditions on the user experience is important for improving the telecommunication services. The QoE for various wireless services including VoIP, video streaming, and web browsing, has been in the epicenter of recent networking activities. The majority of such efforts aim to characterize the user experience, analyzing various types of measurements often in an aggregate manner. This paper proposes the MLQoE, a modular algorithm for user-centric QoE prediction. The MLQoE employs multiple machine learning (ML) algorithms, namely, Artificial Neural Networks, Support Vector Regression machines, Decision Trees, and Gaussian Naive Bayes classifiers, and tunes their hyper-parameters. It uses the Nested Cross Validation (nested CV) protocol for selecting the best classifier and the corresponding best hyper-parameter values and estimates the performance of the final model. The MLQoE is conservative in the performance estimation despite multiple induction of models. The MLQoE is modular, in that, it can be easily extended to include other ML algorithms. The MLQoE selects the ML algorithm that exhibits the best performance and its parameters automatically given the dataset used as input. It uses empirical measurements based on network metrics (e.g., packet loss, delay, and packet interarrival) and subjective opinion scores reported by actual users. This paper extensively evaluates the MLQoE using three unidirectional datasets containing VoIP calls over wireless networks under various network conditions and feedback from subjects (collected in field studies). Moreover, it performs a preliminary analysis to assess the generality of our methodology using bidirectional VoIP and video traces. The MLQoE outperforms several state-of-the-art algorithms, resulting in fairly accurate predictions.

The Forthroid is a location-based system that “augments” physical objects with multimedia information and enables users to receive information about physical objects or request services related to physical objects. It employs computer-vision techniques and Quick Response codes (QR-codes). We have implemented a prototype on Android platforms and evaluated its performance with systems metrics and subjective tests. We discuss our findings and challenges in prototyping on Android OS. The analysis indicates that the network and the server are the main sources of delay, while the CPU load may vary depending on the specific Forthroid operation. The preliminary subjective test results suggest that users tolerate these delays and the offered services can be particularly useful.