Publications

The most common tools to understand perception of food products are hall tests, surveys and observations. However, these approaches require large samples to get reliable results and they are rather costly and time-consuming. Furthermore, they are also highly expert-dependent and rely on the assumption that study participants can express their preferences consciously and explicitly. In our paper, we suggest an electroencephalography- based (EEG) approach to evaluate perceived product similarity in a cross-modal taste-visual task. We tested two potential neurometrics measured from Fz electrode: the amplitude of the N400-like evoked response potentials (ERP) and the power of induced gamma oscillations during 400-600 ms period after visual stimulus presentation. Both metrics showed a strong correlation with the perceived similarity scores at both individual and group levels; however, N400-like amplitude had greater inter-subject variability making it less suitable for practical applications. The results based on the power of induced gamma oscillations (N=18) could be compared to traditional hall-tests (N=200) and may potentially reveal subtle differences in food perception that can not be captured in the hall-tests.

In this paper, we address several aspects of applying classical machine learning algorithms to a regression problem. We compare the predictive power to validate our approach on a data about revenue of a large Russian restaurant chain. We pay special attention to solve two problems: data heterogeneity and a high number of correlated features. We describe methods for considering heterogeneity — observations weighting and estimating models on subsamples. We define a weighting function via Mahalanobis distance in the space of features and show its predictive properties on following methods: ordinary least squares regression, elastic net, support vector regression, and random forest.

This paper investigates the influence of individual mobility on consumer choice in the mobile phone market. The aim is to understand whether consumer mobility influences the switching of an individual to another tariff within one mobile operator. Data were obtained from 1,154 mobile phone users of one of the largest mobile operators in Russia for the period from November 2011 to November 2012. Customers had limited choices in fast and low-cost switching from one operator to another. Hence, we use data of users switching between tariffs within one operator. The data were analyzed by a mixed logit model which is applied to a discrete choice from a variety of disordered alternatives. This solves the restriction of the standard logit model by allowing for random taste variation. We developed three mobility metrics limiting the drawbacks of established metrics. The results show that mobility does influence consumer choices. Tariff plans differ from each other according to the categories of people that use them. Subscriber mobility is one of the features that separate these categories. There are tariffs preferred by highly mobile people, and those that are preferred by people with low mobility. When the individual’s mobility changes, it is highly likely that she will switch to another tariff plan. Mobile operators should take consumer mobility into account when analyzing consumer behavior to increase retention rates.

Understanding neurological mechanisms of motor recovery after stroke is important for selecting appropriate therapeutic and rehabilitation strategies. One of the most widely-used but yet rather controversial MRI predictors is a co-called lesion load on the cortico-spinal tract (CST). This metric corresponds to the overlap between the volumes of the lesion and the cortico-spinal tract which is responsible for conducting neuronal signals that lead to motion generation. In this study we evaluated the potential of the lesion load to explain the motor outcome in a cohort of patients with chronic ischemic stroke. Lesions were automatically identified on structural T1-weighted images using LINDA package. Once lesions are identified, lesion loads on CST were calculated automatically using PALS software package (Ito et al., 2018). Finally, the obtained results were used to classify patients according to their motor outcome using decision tree classifier J48 implemented in WEKA software. However, the classification accuracy was much lower compared to the classification results based on another widely accepted MRI parameter: asymmetry of the fractional anisotropy in the internal capsule of the CST.

The degree of mental attention in childhood and adolescence determines in the future the effectiveness of working memory (ability to store and manipulate information). Attention has been previously found to be related to the prefrontal and parietal areas of the human cortex. But the relationship between attention and white matter properties are still largely unknown. The goal of this study was to identify the relationships between attention and fractional anisotropy (FA) of diffusion MRI in bilateral superior longitudinal fasciculus (in three subdivisions SLF 1- 3), arcuate fasciculus (AF), and corpus callosum (CC) in children and adolescents. Subjects: 14 children (9-11 years) and 13 teenagers (12-15 years). During the experiments participants had to establish a match between the colors on the screen and the colors on the previous slide. The task had six difficulty levels and both performance accuracy (m-score) and reaction time (RT) were measured. There was a positive correlation for m-score and a negative correlation for RT with FA in СС (levels 1-3) in the children's group (p<0.05). On the contrary, when FA increases in the right SLF 3 (level 6), there is a decrease in m-score, and when FA increases in the left SLF 3 and AF, there is an increase in RT at 2,3,4 and 6 levels. In contrast, a decrease in RT with an increase FA of bilateral SLF 3 (level 6) and left AF (level 4) was observed for adolescents, which reflects the redistribution of the roles between fiber tracts with age. FA values of the left (level 2) and right (level 1) SLF 2 negatively correlated with mscore (p <0.05) in the same group. For females (n=13) (regardless the age), there was only a negative correlation for m-score (2,3,5 levels) and the only positive correlation for RT (level 2) with FA of the right SLF 1, left and right SLF 2, in the left SLF 3 and СС (p<0.05). For males (n=13), on the contrary, there were positive correlations between m-score and FA of the СС (1,3,4 levels) and the left SLF 1 (5 level), and inverse correlations between RT and FA for the same fibers of the white matter (1 level) (p<0,05). Interestingly, an increase in FA with age was found in males in all the components of the white matter (p<0.01), except for the СС, and in females, on thecontrary – only in the СС. Further research is needed, taking into account gender, to fully understand the influence of white matter on the development of mental attention.

Patent foramen ovale (PFO) is an important cause of embolic cryptogenic stroke (ECS) in young patients. The main mechanism in this case is paradoxical embolism (PE), the basis for which is a right-to-left (R-L) shunt. Objective: to comparatively characterize patients who have undergone ECS, with and without an R-L shunt, as evidenced by transcranial Doppler with the bubble test (TCD-BT).

The most common tools to understand perception of food products are hall-tests, surveys, observations, etc. However to get reliable results these approaches require large samples, making them costly and time-consuming. Furthermore, they are also highly expert-dependent and rely on the assumption that study participants can express their preferences consciously and explicitly. Here we suggested an EEG-based approach to evaluate perceived product similarity in a cross-modal taste-visual task. Two candidate neurometrics measured from Fz electrode were tested: the amplitude of N430-620 from evoked response potentials (ERP) and the power of induced gamma oscillations during 400-600 ms period after visual stimulus presentation. Both suggested metrics showed a strong correlation with the perceived similarity scores at both individual and group levels, however N430-630 had greater inter-subject variability making it less suitable for practical applications. The results based on the power of induced gamma oscillations (N=18) not only reflected the results from traditional hall-tests (N=200) but also allowed better discrimination between different food products.

Processing of mathematical operations and solving numerical tasks implicate a distributed set of brain regions. These regions include the superior and inferior parietal lobules that underlie numerical processing such as size judgments, and additional prefrontal regions that are needed for formal mathematical operations such as addition, subtraction and multiplication [Arsalidou, Taylor, 2011]. Critically, little is known about the connectivity between these regions and the association between math performance and the anatomical structure of white matter tracts. The present study investigates connectivity and white matter tracks associated with networks related to math performance: arcuate fasciculus (AF) and superior longitudinal fasciculus (SLF). Participants performed a computerized task with mathematical operations (addition, subtraction, multiplication, and division) with three levels of difficulty; accuracy and reaction time were recorded. Diffusion tensor imagining (DTI) recordings provided indices on fractional anisotropy (FA) — a measure of the direction of white matter tracks in the brain. The relation between FA and math performance scores is reported.

This paper is an empirical study of the changing nature of the dependence of fundamental factors on the stock market index, which is the trend identified earlier in the Russian stock market. We empirically test the impact of daily values of fundamental factors on the MOEX Russia Index from 2003 to 2018. The analysis of the ARIMA-GARCH (1,1) model with a rolling window reveals that the change in the power and direction of the influence of the fundamental factors on the Russian stock market persists. The Quandt-Andrews breakpoint test and Bai-Perron test identify the number and likely location of structural breaks. We find multiple breaks probably associated with the dramatic falls of the stock market index. The results of the regression models over the different regimes, defined by the structural breaks, can vary markedly over time. This research is of value in macroeconomic forecasting and in the investment strategy development

This paper studies how the influence of the fundamental factors on the Russian stock market changes retrospectively. We empirically test the impact of daily values of fundamental factors (indexes of foreign stock markets, oil price, exchange rate and interest rates in Russia and the USA) on the MOEX Russia Index over long time interval from 2003 to 2018. The analysis of the ARIMA-GARCH (1,1) model with a rolling window reveals the changes in the power and direction of the influence of the fundamental factors which are probably caused by the structural instability revealed earlier in Russia and other stock markets. The Quandt-Andrews breakpoint test and Bai-Perron test identify the number and likely location of the structural breaks. We find multiple breaks probably associated with dramatic falls in the stock market index, for example with the significant falls of the then MICEX index in the spring of 2006 and the global financial crisis of 2008-2009. The results of the regression models over the different regimes, defined by the structural breaks, can vary markedly over time.

This paper describes the results of research project on optimal pricing for LLC "Perm Local Rail Company". In this study we propose a regression tree based approach for estimation of demand function for local rail tickets considering high degree of demand heterogeneity by various trip directions and the goals of travel. Employing detailed data on ticket sales for 5 years we estimate the parameters of demand function and reveal the significant variation in price elasticity of demand. While in average the demand is elastic by price, near a quarter of trips is characterized by weakly elastic demand. Lower elasticity of demand is correlated with lower degree of competition with other transport and inflexible frequency of travel.

  In many practical tasks it is needed to estimates effect of a treatemnt on individual level. For example in medicine it is essential to determine the patients that would benifit from a certain medicament. In marketing knowning the persons that are likely to buy a new product would reduce the amount of spam. In this chapter we review the methods to estimate individulize treatment effect from a randomized trial, i.e., an experiment when a part of individuals recieves a new treatment, while the others does not. Finally, it is shown that new efficient methods are needed in this domain.

In this paper, we analyze a new approach for demand prediction in retail. One of the signicant gaps in demand prediction by machine learning methods is the unaccounted sales data censorship. Econometric approaches to modeling censored demand are used to obtain consistent and unbiased estimates of parameters. These approaches can also be transferred to different classes of machine learning models to reduce the prediction error of sales volume. In this study we build two ensemble models to predict demand with and without demand censorship, aggregating predictions for machine learning methods such as Linear regression, Ridge regression, LASSO and Random forest. Having estimated the predictive properties of both models, we test the best predictive power of the models with accounting for the censored nature of demand.

Background and Purpose Despite the continuing efforts in multimodal assessment of the motor system after stroke, conclusive findings on the complementarity of functional and structural metrics of the corticospinal tract (CST) integrity and the role of the contralesional hemisphere are still missing. The aim of this work was to find the best combination of the motor system parameters, allowing classification of patients into three predefined groups of upper limb motor recovery. Methods 35 chronic ischemic stroke patients (47 [26–66] y.o., 29 [6–58] months post-stroke) with only supratentorial lesion and unilateral upper extremity weakness were enrolled. Patients were divided into three groups depending on the upper limb motor recovery. Non-parametric statistical tests and regression analysis were used to investigate the relationships among structural and functional motor system parameters, probed by diffusion tensor imaging (DTI) and transcranial magnetic stimulation (TMS). In addition, stratification rules were tested, using a decision tree classifier to identify parameters explaining motor recovery. Results Fractional anisotropy (FA) ratio in the internal capsule (IC) and absence/presence of motor evoked potentials (MEPs), were equally discriminative of the worst motor outcome group (96% accuracy). MEP presence diverged for two investigated hand muscles. Concurrently, for the three recovery groups’ classification, the best parameter combination was: IC FA ratio and Fréchet distance between the contralesional and ipsilesional CST FA profiles (91% accuracy). No other metrics had any additional value for patients’ classification. Conclusions This study demonstrates that IC FA ratio and MEPs absence are equally important markers for poor recovery. Importantly, we found that MEPs should be controlled in more than one hand muscle. Finally, we show that better separation between different motor recovery groups may be achieved when considering the whole CST FA profile.

Proceedings of the Fifth Workshop on Experimental Economics and Machine Learning at the National Research Univeristy Higher School of Economics co-located with the Seventh International Conference on Applied Research in Economics (iCare7)

In this research we analyze a new approach for prediction of demand. In the studied market of performing arts the observed demand is limited by capacity of the house. Then one needs to account for demand censorship to obtain unbiased estimates of demand function parameters. The presence of consumer segments with different purposes of going to the theater and willingness‐to‐pay for performance and ticket characteristics causes a heterogeneity in theater demand. We propose an estimator for prediction of demand that accounts for both demand censorship and preferences heterogeneity. The estimator is based on the idea of classification and regression trees and bagging prediction aggregation extended for prediction of censored data. Our algorithm predicts and combines predictions for both discrete and continuous parts of censored data. We show that our estimator performs better in terms of prediction accuracy compared with estimators which account either for censorship or heterogeneity only. The proposed approach is helpful for finding product segments and optimal price setting.

Schizophrenia is a chronic debilitating disease. Sleep disturbances associated with a reduction in spindles are observed as warning signs prior to the first psychotic episode. Every spindle is a short-lasting (~0.5 s) set of bioelectric sinusoidal waves at the frequency of 10-16 Hz generated within the thalamus. Sleep spindles, easily identifiable in a scalp electroencephalogram, occur hundreds of times during sleep and are implicated in cognition like memory processes. For this reason, spindles are seen as an electrobiomarker of the quality of sleep and cognitive performance. In patients at high risk of psychotic transition, the density (number/time unit) of spindles is reduced. The underlying mechanisms of this change are unknown. Glutamate-mediated neurotransmission in the thalamus plays a key role in the generation of spindles and the etiology of schizophrenia. Therefore, we tested the hypothesis that a reduced function of glutamate receptors at the thalamic level is involved in the psychosis-related reduction in spindles. Using cell-to-network neurophysiological methods in sleeping rats, we demonstrate that systemic administration of the NMDA glutamate receptor antagonist, ketamine, significantly decreases spindle density. This effect is consistently prevented by the widely used antipsychotic drug, clozapine. These original findings support the hypothesis of the involvement of a reduced function of NMDA glutamate receptors in the sleep spindle deficit observed in psychosis-related disorders. The present findings lay the foundation for the development of innovative therapies aimed at preventing psychotic, bipolar, and depressive disorders.