No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
How to predict choice using eye-movements data?
Abstract
In recent decades, eye-movement detection technology has improved significantly, and eye-trackers are available not only as standalone research tools but also as computer peripherals. This rapid spread gives further opportunities to measure the eye-movements of participants. The current paper provides classification models for the prediction of food choice and selects the best one. Four choice sets were presented to 112 volunteered participants, each choice set consisting of four different choice tasks, resulting in altogether sixteen choice tasks. The choice sets followed the 2, 4, 6 and 8-alternative forced-choice paradigm. Tobii X2-60 eye-tracker and Tobii Studio software were used to capture and export gazing data, respectively. After variable filtering, thirteen classification models were elaborated and tested; moreover, eight performance parameters were computed. The models were compared based on the performance parameters using the sum of ranking differences algorithm. The algorithm ranks and groups the models by comparing the ranks of their performance metrics to a predefined gold standard. Techniques based on decision trees were superior in all cases, regardless of the choice tasks and food product categories. Among the classifiers, Quinlan's C4.5 and cost-sensitive decision trees proved to be the best-performing ones. Future studies should focus on the fine-tuning of these models as well as their applications with mobile eye-trackers.
... By tracking the movement and focus of the eyes, researchers gain insights into what consumers pay attention to, how long they focus on different elements, and how their visual attention influences their behavior [1]. Furthermore, eye-tracking techniques offer a promising avenue for objectively assessing consumers' visual attention to products [2,3]. Both in Google scholar and Scopus, the key word of "eye-tracking" has shown a clear upward trend during the past five years. ...
... (2) How do the appearance and defects of fruits influence their purchase decisions? (3) What are the factors that may alter consumer perception and acceptance of visually suboptimal but nutritionally adequate fruits, aiding in reducing food wastage? ...
... (Tobii Pro AB, Danderyd, Sweden) software. As reported by the time to first fixation, fixation duration and dwell count are critical metrics for understanding consumer behavior in food choice tasks, making them suitable measures for analyzing visual attention in this study [3]. The following eye-tracking parameters were extracted and used for the data analysis [4]: Independent samples t-test was used to compare the means of the different eye-tracking parameters of participants coming from the two countries (CHN and HUN). ...
Using eye-tracking technology, the proposed study investigates how customers visually evaluate apples varieties and apple defects and how these evaluations affect their purchasing decisions. Three aspects were examined in this study: apple variety, defect severity, and cultural background. Idared, Golden Delicious Yellow, and Golden Delicious Green apple varieties with increasing degrees of bruising were shown to Chinese and Hungarian participants. The findings show that apple variety had no significant effect on gaze patterns, whereas cultural background had a considerable impact on visual attention measures. The most important element in grabbing and retaining customer attention was the severity of the defect, which was measured by area. The “Threshold of Rejection”, which characterizes consumer tolerance for Apple defects, is introduced in the study. Furthermore, a polynomial regression model was created to predict the probability of repurchasing an apple depending on its visual quality (level of bruising). These results provide useful information for marketing plans, quality assurance, and comprehending customer behavior in the fresh produce sector.
... Tracking Devices or Sensors are tools used to gather data about how people interact with things in AV. Eye-tracking cameras follow where people look at, helping understand how people choose food or products (Gere et al., 2021b). Examples include Tobii AB (Danderyd Municipality, Sweden) and iMotions A/S (Copenhagen, Denmark). ...
... By selecting various food items, researchers can assess the versatility of AV technology across culinary experiences, enriching sensory science and consumer insights. Virtual and augmented reality technologies show potential in sensory science, especially in studying meal choices and testing usability in a virtual reality food court (Chai et al., 2022;Gere et al., 2021b). Moreover, these technologies relate to consumer consciousness in multisensory extended reality, emphasizing their impact on perception and psychology (Petit et al., 2022). ...
Augmented Virtuality (AV) is a concept that merges components of Augmented Reality (AR) and Virtual Reality (VR), incorporating real elements into a virtual environment. This review analyses the influence of AV technology on sensory science and consumer behaviour, with the potential to improve product evaluation through sensory analysis. The objective is to develop immersive sensory environments that closely resemble real-world scenarios, offering accurate insights into consumer perceptions and preferences. Participants will be able to observe genuine food products within the virtual environment. Through the utilization of a multidisciplinary approach, the analysis explores the point at which technology and human senses intersect, revealing new and unique understandings of decision-making processes. This enhances comprehension of consumer choices and behaviour in virtual environments, providing practical uses for industries navigating the ever-changing nature of augmented virtuality. This review demonstrates that the integration of AV elements in sensory science can have a substantial influence.
... Second, via comparison of the predictors by a versatile non-parametric statistical tool: the sum of ranking differences and comparison of ranks to random numbers (SRD-CRRN) method augmented by one-way analysis of variance (ANOVA) with post hoc Bonferroni pair-wise tests. [141][142][143][144] 3.1. The visual comparison of RCCS predictors for C a chemical shis 3.1.1. ...
... These mean SRD score values correspond to the positions of the stick in Fig. 7. Which of the ve methods are signicantly better or worse can be checked by a Bonferroni post hoc test. 144 The Bonferroni test provides a grouping pattern of the methods as shown in Table 2. In this example, the chosen ve methods form four homogenous groups named G1, G2, G3 and G4. ...
In studying secondary structural propensities of proteins by nuclear magnetic resonance (NMR) spectroscopy, secondary chemical shifts (SCSs) serve as the primary atomic scale observables. For SCS calculation, the selection of an appropriate random coil chemical shift (RCCS) dataset is a crucial step, especially when investigating intrinsically disordered proteins (IDPs). The scientific literature is abundant in such datasets, however, the effect of choosing one over all the others in a concrete application has not yet been studied thoroughly and systematically. Hereby, we review the available RCCS prediction methods and to compare them, we conduct statistical inference by means of the nonparametric sum of ranking differences and comparison of ranks to random numbers (SRD-CRRN) method. We try to find the RCCS predictors best representing the general consensus regarding secondary structural propensities. The existence and the magnitude of resulting differences on secondary structure determination under varying sample conditions (temperature, pH) are demonstrated and discussed for globular proteins and especially IDPs.
... Finally, the current visualisations mostly focused on general data exploration. Studies increasingly make use of machine learning and deep learning methods, for example to predict choice from patterns of eye movements (Gere et al. 2021;Unger et al. 2023). Future research should therefore establish which types of visualisations are most effective for exploratory data analysis in this context. ...
Eye movements have a spatial (where people look), but also a temporal (when people look) component. Various types of visualizations have been proposed that take this spatio-temporal nature of the data into account, but it is unclear how well each one can be interpreted and whether such interpretation depends on the question asked about the data or the nature of the dataset that is being visualised. In this study, four spatio-temporal visualization techniques for eye movements (chord diagram, scan path, scarf plot, space-time cube) were compared in a user study. Participants ( N = 25 ) answered three questions (what region first, what region most, which regions most between) about each visualization, which was based on two types of datasets (eye movements towards adverts, eye movements towards pairs of gambles). Accuracy of the answers depended on a combination of the dataset, the question that needed to answered, and the type of visualization. For most questions, the scan path, which did not use area of interest (AOI) information, resulted in lower accuracy than the other graphs. This suggests that AOIs improve the information conveyed by graphs. No effects of experience with reading graphs (for work or not for work) or education on accuracy of the answer was found. The results therefore suggest that there is no single best visualisation of the spatio-temporal aspects of eye movements. When visualising eye movement data, a user study may therefore be beneficial to determine the optimal visualization of the dataset and research question at hand.
Graphical abstract
... Further statistical analysis was performed on the eye-tracking parameters extracted from the eye-tracker measurements, which were as follows (Gere et al., 2021): The analysis of the above parameters was also performed using STATISTICA v.10 (Statsoft Inc., Tulsa, Oklahoma, USA), and analysis of variance (ANOVA) was used. In order to analyze the eye-tracking parameters, AOIs (Areas of Interest) were first assigned for each product. ...
In our study, using a combination of eye-tracking parameter analysis and the van Westendorp method, we investigate whether participants pay more attention to products that they perceive as more expensive or to those that they prefer in the ranking process. The experiment involved 50 participants, a questionnaire with ranking and pricing tasks, and an eye-tracking measurement. Three wine varieties (Irsai Olivér, Rosé and Merlot-Shiraz) and three different label alternatives were tested. When comparing the results of the ranking and the pricing tasks, the product that is considered more expensive is not always the one that is most appealing to the participants. If we compare the results from the analysis of the eye-tracking parameters and the pricing, we can say that in all cases the labels that received the most visual attention were those that were priced more expensively by the participants.
... Fenko et al. (2018) reported that the visual attention to health labels was a poor predictor of healthy choices. However, Gere et al. (2021) suggested that applying statistical choice prediction models to gazing data was a suitable approach to predict food choice in different product categories. Different authors have reported that during a decision-making task, higher fixation counts, and longer fixation duration were related to the alternative selected (Danner et al., 2016;van der Laan et al., 2015). ...
... Nevertheless, variations in RRV snack anticipated weight or fat mass (FM) decreased, indicating that dietary therapies that either reduce or encourage RRV snack decreases produced a greater reduction in obese. Gere et al. [11] provided a classification approaches for food preference prediction. Thirteen classification models were developed and evaluated after variable filtering. ...
Food choice motives (i.e., mood, health, natural content, convenience, sensory appeal, price, familiarities, ethical concerns, and weight control) have an important role in transforming the current food system to ensure the healthiness of people and the sustainability of the world. Researchers from several domains have presented several models addressing issues influencing food choice over the years. However, a multidisciplinary approach is required to better understand how various aspects interact with one another during the decision-making procedure. In this paper, four Deep Learning (DL) models and one Machine Learning (ML) model are utilized to predict the weight in pounds based on food choices. The Long Short-Term Memory (LSTM) model, stacked-LSTM model, Conventional Neural Network (CNN) model, and CNN-LSTM model are the used deep learning models. While the applied ML model is the K-Nearest Neighbor (KNN) regressor. The efficiency of the proposed model was determined based on the error rate obtained from the experimental results. The findings indicated that Mean Absolute Error (MAE) is 0.0087, the Mean Square Error (MSE) is 0.00011, the Median Absolute Error (MedAE) is 0.006, the Root Mean Square Error (RMSE) is 0.011, and the Mean Absolute Percentage Error (MAPE) is 21. Therefore, the results demonstrated that the stacked LSTM achieved improved results compared with the LSTM, CNN, CNN-LSTM, and KNN regressor.
... In the later stage of free observation, the main regions of interest (i.e., ROI I is the square region of the left food image and ROI II is the square region of the right food image) were the square regions for each food image. We calculated five basic eye movement measures to examine the later-stage, conscious food-related attentional bias: (i) time to first fixation is the duration between the appearance of the stimulus and the time that the participant first fixated on the ROI (i.e., ROI I or ROI II) in the free-viewing task after the decision task (Gere et al., 2021); (ii) first fixation duration is the time spent on the first fixation for each food image item; (iii) first fixation direction bias (%) is defined as the percentage of participants who looked at the high-calorie food for the first time in all trials (Akcay et al., 2022); (iv) mean fixation duration is the mean of the average fixation time for the high-or low-calorie food images in all trials; (v) total fixation duration is the sum time of all fixations for each food image in a single trial (Rayner, 2009). In addition, the participant was excluded if the effective data was less than 75% since this meant that the participants did not follow the instructions properly. ...
Food-related attentional bias has been studied for many years, yet the time course of attentional bias is not well characterized. Probe detection and Stroop paradigms are commonly utilized to examine food-related attentional bias, however, these methods are relatively rough in reflecting attentional processing. Thus, we used a modified food-house task combined with eye-tracking to investigate restrained and unrestrained eaters’ food-related attentional bias in different time courses. Saccade trajectory deviations and fixation durations were collected as eye movement measures to examine unconscious detection bias in the early stage of attentional processing and conscious maintenance bias in the later stage. An approach-avoidance conflict towards low-calorie food cues was found in restrained eaters, showing that food-related attentional bias changes with the different time courses. The saccade curvature demonstrated an early-stage attentional bias towards low-calorie foods in restrained eaters, whereas fixation measurements suggested a later-stage attentional avoidance of low-calorie foods in restrained eaters. The processing accumulation of food cues in human consciousness can probably explain the results. With the increase of the priming effect of high-calorie foods, attentional bias towards low-calorie foods disappears. In this study, saccade curvature was confirmed to be useful for directly revealing early-stage unconscious food-related attentional processing, and the role of time courses in attention allocation was also demonstrated.
... Due to its flexibility and ease of use, SRD has been used in different fields of science such as eye-tracking [10]; food science [11]; column selection in chromatography [12,13]; variable selection [14]; ordering and grouping octanol-water (logP) partition coefficient determination methods [15,16]; selection of edible insects based on nutritional composition [17]; outlier detection in multivariate calibration [18]; non-parametric ranking of QSAR models [19]; comparison of ensemble learning models [20]; comparison of tea grade identification using electronic tongue data [21]; testing the outer consistency of novel similarity indices [22]; and even ranking of sportsmen [23], just to name a few. ...
Predicting the success of National Football League drafts has always been an exciting issue for the teams, fans and even for scientists. Among the numerous approaches, one of the best techniques is to ask the opinion of sport experts, who have the knowledge and past experiences to rate the drafts of the teams. When asking a set of sport experts to evaluate the performances of teams, a multicriteria decision making problem arises unavoidably. The current paper uses the draft evaluations of the 32 NFL teams given by 18 experts: a novel multicriteria decision making tool has been applied: the sum of ranking differences (SRD). We introduce a quick and easy-to-follow approach on how to evaluate the performance of the teams and the experts at the same time. Our results on the 2021 NFL draft data indicate that Green Bay Packers has the most promising drafts for 2021, while the experts have been grouped into three distinct groups based on the distance to the hypothetical best evaluation. Even the coding options can be tailored according to the experts’ opinions. Statistically correct (pairwise or group) comparisons can be made using analysis of variance (ANOVA). A comparison to TOPSIS ranking revealed that SRD gives a more objective ranking due to the lack of predefined weights.
Playing ethnic music in restaurants increases consumer experience. Studies show, furthermore, that ethnic congruence of music and food affects food selection but not the liking of customers. An eye-tracking study was completed with 104 participants to uncover if there is an effect of ethnic music on selecting ethnic foods. German, Hungarian, Italian, and Spanish ethnic music was played while participants choose congruent starters, main dishes, and desserts. Results show that visual attention decreased when any background music was played. However, when played, the highest visual attention was recorded during Spanish music. Similarly, the most visual attention was recorded on Spanish dishes. Food choice frequencies showed no differences among the four nations. However, after aggregating German-Hungarian and Italian-Spanish music and dishes, it turned out that participants chose congruent music and food. Choice predictions were also completed on data with and without ethnic music. The performance of prediction models significantly increased when music was played. These findings highlight a clear link between music and food choices, and that music helped participants complete their choices and decide faster.
This study aimed to select the feature genes of hepatocellular carcinoma (HCC) with the Fisher score algorithm and to identify hub genes with the Maximal Clique Centrality (MCC) algorithm. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis was performed to examine the enrichment of terms. Gene set enrichment analysis (GSEA) was used to identify the classes of genes that are overrepresented. Following the construction of a protein-protein interaction network with the feature genes, hub genes were identified with the MCC algorithm. The Kaplan–Meier plotter was utilized to assess the prognosis of patients based on expression of the hub genes. The feature genes were closely associated with cancer and the cell cycle, as revealed by GO, KEGG and GSEA enrichment analyses. Survival analysis showed that the overexpression of the Fisher score–selected hub genes was associated with decreased survival time (P < 0.05). Weighted gene co-expression network analysis (WGCNA), Lasso, ReliefF and random forest were used for comparison with the Fisher score algorithm. The comparison among these approaches showed that the Fisher score algorithm is superior to the Lasso and ReliefF algorithms in terms of hub gene identification and has similar performance to the WGCNA and random forest algorithms. Our results demonstrated that the Fisher score followed by the application of the MCC algorithm can accurately identify hub genes in HCC.
Machine learning classification algorithms are widely used for the prediction and classification of the different properties of molecules such as toxicity or biological activity. the prediction of toxic vs. non-toxic molecules is important due to testing on living animals, which has ethical and cost drawbacks as well. The quality of classification models can be determined with several performance parameters. which often give conflicting results. In this study, we performed a multi-level comparison with the use of different performance metrics and machine learning classification methods. Well-established and standardized protocols for the machine learning tasks were used in each case. The comparison was applied to three datasets (acute and aquatic toxicities) and the robust, yet sensitive, sum of ranking differences (SRD) and analysis of variance (ANOVA) were applied for evaluation. The effect of dataset composition (balanced vs. imbalanced) and 2-class vs. multiclass classification scenarios was also studied. Most of the performance metrics are sensitive to dataset composition, especially in 2-class classification problems. The optimal machine learning algorithm also depends significantly on the composition of the dataset.
Purpose
The relation between preference and the gaze for the test foods under unconsciousness using the eye-tracking system was investigated.
Methods
Participants consisted of 37 healthy volunteers. Test foods were steamed rice 150 g, grilled salmon approximately 45 g and slice cooked squash 60 g, all of which were served on a tray. Foods forms were regular food, chopped food, and blended food. After attached to the eye tracker, participants watched the each dish arranged in front of them freely for 10 s. And they ate test foods freely within 10 min. The gazing point was measured for 10 s from the time when the food was ordered and just before the eating. Preference levels were interviewed. The number of gaze point fixations and the total gaze point fixation time of the viewpoint during 10 s just before eating were analyzed. The analysis items were (1) the total number of gaze point fixations (2) the total gaze point fixation time (3) the amount of food intake and (4) the preference level details.
Results
For foods with higher preference levels, the number of gaze point fixations increased significantly and the total gaze point fixation time significantly increased. In both groups, maximum food intake was observed for food forms with a high preference level. Most of the participants’ selected regular foods as their most preferred food form among the food forms.
Conclusions
The results suggested that subjects gazed at regular food which had high preference level.
This study investigates the relationship between gazing behavior and choice decision in multialternative forced choice tasks, focusing on the consistency across different food product groups including apple, beer, bread, chocolate, instant soup, salad, sausage and soft drink. Each choice task consisted of pictures of four alternatives, similar in familiarity and liking ratings, of the corresponding product group. A Tobii T60 eye-tracker was used to present the stimuli and to analyze the gazing behavior of 59 participants during decision-making.
The results showed strong correlations between choice and gazing behavior, in forms of more fixation counts, longer total dwell duration and more dwell counts on the chosen alternative. No correlations for first fixation, time to first fixation and first fixation duration were observed. These results were consistent across the eight tested product groups.
Classification in data mining has gained a lot of importance in literature and it has a great deal of application areas from medicine to astronomy, from banking to text classification.. It can be described as supervised learning algorithm as it assigns class labels to data objects based on the relationship between the data items with a pre-defined class label. The classification techniques are help to learn a model from a set of training data and to classify a test data well into one of the classes. This research is related to the study of the existing classification algorithm and their comparative in terms of speed, accuracy, scalability and other issues which in turn would help other researchers in studying the existing algorithms as well as developing innovative algorithms for applications or requirements which are not available.
Current inductive machine learning algorithms typically use greedy search with limited lookahead. This prevents them to detect significant conditional dependencies between the attributes that describe training objects. Instead of myopic impurity functions and lookahead, we propose to use RELIEFF, an extension of RELIEF developed by Kira and Rendell [10, 11], for heuristic guidance of inductive learning algorithms. We have reimplemented Assistant, a system for top down induction of decision trees, using RELIEFF as an estimator of attributes at each selection step. The algorithm is tested on several artificial and several real world problems and the results are compared with some other well known machine learning algorithms. Excellent results on artificial data sets and two real world problems show the advantage of the presented approach to inductive learning.
Supervised classification is one of the tasks most frequently carried out by so-called Intelligent Systems. Thus, a large
number of techniques have been developed based on Artificial Intelligence (Logic-based techniques, Perceptron-based techniques)
and Statistics (Bayesian Networks, Instance-based techniques). The goal of supervised learning is to build a concise model
of the distribution of class labels in terms of predictor features. The resulting classifier is then used to assign class
labels to the testing instances where the values of the predictor features are known, but the value of the class label is
unknown. This paper describes various classification algorithms and the recent attempt for improving classification accuracy—ensembles
of classifiers.
Fisher score is one of the most widely used supervised feature selection
methods. However, it selects each feature independently according to their
scores under the Fisher criterion, which leads to a suboptimal subset of
features. In this paper, we present a generalized Fisher score to jointly
select features. It aims at finding an subset of features, which maximize the
lower bound of traditional Fisher score. The resulting feature selection
problem is a mixed integer programming, which can be reformulated as a
quadratically constrained linear programming (QCLP). It is solved by cutting
plane algorithm, in each iteration of which a multiple kernel learning problem
is solved alternatively by multivariate ridge regression and projected gradient
descent. Experiments on benchmark data sets indicate that the proposed method
outperforms Fisher score as well as many other state-of-the-art feature
selection methods.
Most organisms facing a choice between multiple stimuli will look repeatedly at them, presumably implementing a comparison process between the items' values. Little is known about the nature of the comparison process in value-based decision-making or about the role of visual fixations in this process. We created a computational model of value-based binary choice in which fixations guide the comparison process and tested it on humans using eye-tracking. We found that the model can quantitatively explain complex relationships between fixation patterns and choices, as well as several fixation-driven decision biases.
We develop a reporting guideline for eye-tracking research in the behavioral sciences. To this end, we coded 215 articles on behavioral decision-making published between 2009 and 2017 and extracted a list of reported items. The coded articles were from a broad range of disciplines linked to judgement and decision making, such as cognitive science, marketing, economics, developmental research, vision research, and human–computer interaction. We then had a panel of eye-tracking experts rate the necessity of each item for reproducing a reported study. From these two sources, we generated a guideline containing 31 items that are judged as 'necessary' by the majority of experts for reproducing an eye-tracking study. None of the 215 coded articles report all identified items and approximately 70 percent of the articles report less than 50 percent of the 'necessary' items. We provide the data and list of recommendations as a hands-on shiny app to allow for an easy adoption of the proposed reporting guideline to improve transparency and reproducibility in eye-tracking research.
Well-structured stimuli presentation is essential in eye-tracking research to test predefined hypotheses reliably and to conduct relevant gazing behavior studies. Several bottom-up factors associated with stimuli presentation (such as stimuli orientation, size etc.) can influence gazing behavior. However, only a small number of scientific papers address these factors in a sensory and consumer science context and thus provide guidance to practitioners. The two presented eye-tracking studies on food images aimed at evaluating the effect of the bottom-up factors stimulus size, background of the picture, orientation of food product presentation, the evaluated products and the number of alternatives. Significant effects of product group were found in the case of all eye-movement parameters except time to first fixation and first fixation duration. In contrary, orientation significantly influenced only the time to first fixation and first fixation duration parameters. Stimulus size significantly increased fixation and dwell count, while background showed no significant effects. Furthermore, significant relationships were found between the number of presented images and eye-movement and decision time. Less time was needed in 2AFC (alternative forced choice test), 3AFC and 4AFC and significantly more time was needed to choose one alternative out of 7AFC and 8AFC. The results of the two studies show that the investigated bottom-up factors can significantly influence gazing behavior, and therefore need to be carefully considered when planning or comparing results of eye-tracking experiments.
Determining the key parameters driving attention and choice at the point of sale is a challenging task. To address this challenge, we performed two studies employing eye-tracking (ET) as a methodological tool when varying the visual marketing stimuli in a lab-experimental setting and in real supermarket shelf, and thus, facing an important gap in the current body of literature – the need to reconcile ET results from lab and field studies.
The first study was conducted in lab settings and explored in a controlled manner the top-down (goal-directed) vs. bottom-up (stimulus-driven) mechanisms of attention and choice. The second study took a step further in investigating these mechanisms in real life settings, namely a supermarket shelf. In both studies the same assortment context was presented (i.e. eight products, four flavours of two brands each). The products varied on their level of healthfulness (i.e. nutrient profile) which was explicitly communicated with nutrition labelling formats displayed front of pack. Participants were asked to select either the healthiest product or a product on their preference (lab settings), and a product of their preference (in-store settings). Fixation duration, number of fixations, and the consumer's choice was recorded.
The results show that Brand and Product flavour are leading criteria in driving attention and choice, i.e. the stronger brand and best selected product received higher number of fixations. The shopping goal and label formats also contributed to variation in observed patterns. Brand placement in combination with brand strength had a significant impact in the retail environment. Current outcomes demonstrate the potential of eye-tracking in consumer research, from lab to supermarket shelf. The advanced understanding we offer in attention patterns and consequent decision opens promising avenues in successfully applying marketing strategies to navigate consumers’ attention and choice.
The old debate is revived: Definite differences can be observed in suggestions of estimation for prediction performances of models and for validation variants according to the various scientific disciplines. However, the best and/or recommended practice for the same data set cannot be dependent on the field of usage. Fortunately, there is a method comparison algorithm, which can rank and group the validation variants; its combination with variance analysis will reveal whether the differences are significant or merely the play of random errors. Therefore, three case studies have been selected carefully to reveal similarities and differences in validation variants. The case studies illustrate the different significance of these variants well. In special circumstances, any of the influential factors for validation variants can exert significant influence on evaluation by sums of (absolute) ranking differences (SRDs): stratified (contiguous block) or repeated Monte Carlo resampling and how many times the data set is split (5‐7‐10). The optimal validation variant should be determined individually again and again. A random resampling with sevenfold cross‐validations seems to be a good compromise to diminish the bias and variance alike. If the data structure is unknown, a randomization of rows is suggested before SRD analysis. On the other hand, the differences in classifiers, validation schemes, and models proved to be always significant, and even subtle differences can be detected reliably using SRD and analysis of variance (ANOVA).
This chapter provides a general overview of eye-tracking techniques and their applications in consumer research with a focus on the food area. Firstly, the scientific approaches leading to the development of eye-trackers are described, followed by a review of the principles and technical solutions of measuring gazing behavior. After a description of the factors influencing gaze behavior and a discussion of the relation between gaze, choice and decision making we present applications of eye-tracking in the fields of packaging, label and menu design, in-store consumer behavior, emotional responses and eating disorders. Finally, we discuss a case study involving the use of eye-tracking for studying consumer food choice in more detail.
Feature selection plays a critical role in data mining, driven by increasing feature dimensionality in target problems and growing interest in advanced but computationally expensive methodologies able to model complex associations. Specifically, there is a need for feature selection methods that are computationally efficient, yet sensitive to complex patterns of association, e.g. interactions, so that informative features are not mistakenly eliminated prior to downstream modeling. This paper focuses on Relief-based algorithms (RBAs), a unique family of filter-style feature selection algorithms that strike an effective balance between these objectives while flexibly adapting to various data characteristics, e.g. classification vs. regression. First, this work broadly examines types of feature selection and defines RBAs within that context. Next, we introduce the original Relief algorithm and associated concepts, emphasizing the intuition behind how it works, how feature weights generated by the algorithm can be interpreted, and why it is sensitive to feature interactions without evaluating combinations of features. Lastly, we include an expansive review of RBA methodological research beyond Relief and its popular descendant, ReliefF. In particular, we characterize branches of RBA research, and provide comparative summaries of RBA algorithms including contributions, strategies, functionality, time complexity, adaptation to key data characteristics, and software availability.
Visual attention plays an active role in food choice. During eye-tracking, several gazing behavior parameters are measured along with the consumer’s choice. In this study, a Tobii T60 eye-tracker was used to record the gazing behavior of 59 participants during multi-alternative choice tasks (4AFC) in which pictures of six different food product groups (apples, salads, instant soups, sausages, soft drinks and beers) were presented on the eye tracker screen. The aim of this study was to investigate the relationship between gazing parameters and choice and to create prediction models based on gazing data. Furthermore, we aimed to search for the best model and to propose a workflow. The applied thirteen statistical models showed strong relationships between gazing behavior and choice and gave accurate predictions for choice. sum of ranking differences method was used to rank the prediction models based on ten performance indicators. Iterative Dichotomiser 3 algorithm, Quinlan’s C4.5 decision tree algorithm and k-Nearest Neighbour’s algorithm showed the best performances in the cases of the separate product groups. After merging the data sets, Iterative Dichotomiser 3 algorithm showed clearly the best performance to describe the connection between visual attention and food choice.
People often purchase food products on impulse and their visual impression of such products plays an important role in impulse buying. Consumers are also likely to buy food items based on the images as displayed on mobile devices like smartphones. Food-service and dining industries have therefore begun to pay closer attention to improving the visual impression of the foods they offer. This study focused on determining whether participants’ visual attention directed toward food-item images can vary depending on the background saliency. Differences in patterns of visual attention with respect to food-item images between North American and Chinese participants were also compared. During the time participants were looking at pictures of food items with varying backgrounds in the absence of a particular task, their eye movements were traced with an eye-tracker. As background contexts such as table setting and decoration became more salient, participants’ visual attention toward the food items decreased. Chines participants also looked at food items significantly later than American counterparts, implying that Chinese participants were relatively more influenced by background contexts. In conclusion, our findings provide empirical evidence that background context and culture can affect participants’ visual attention while they are freely looking at pictures of food items.
Supervised machine learning is the search for algorithms that reason from externally supplied instances to produce general hypotheses, which then make predictions about future instances. In other words, the goal of supervised learning is to build a concise model of the distribution of class labels in terms of predictor features. The resulting classifier is then used to assign class labels to the testing instances where the values of the predictor features are known, but the value of the class label is unknown. This paper describes various supervised machine learning classification techniques. Of course, a single article cannot be a complete review of all supervised machine learning classification algorithms (also known induction classification algorithms), yet we hope that the references cited will cover the major theoretical issues, guiding the researcher in interesting research directions and suggesting possible bias combinations that have yet to be explored.
This study investigated the effect of food color on gazing behavior using eye-tracking technology and the correlation between gazing behavior and choice decision. Tobii T60 eye-tracker was used for analyzing the gazing behavior of consumers. Images of three different food products with three different colors each (yellow, green, pink) were used as stimuli. Seventy-three subjects were recruited; color blind individuals were excluded from the test. After the eye tracking procedure, the test persons had to decide which sample they preferred. Results show that the colors of the used food products significantly affected the gazing behavior and the choice. Fixation count and visit duration correlated significantly in a positive way with choice rate. This insight highlights the importance of visual attraction for the choosing behavior and it might open the chance to predict choice behavior measuring gazing behavior.
Significant progress has been achieved since the introduction of the new similarity measure: the sum of
absolute ranking differences (SRDs) [TrAC — Trends in Anal. Chem. 29 (2010) 101–109]. Empirical evidences
were accumulated about scaling, selection of the reference (benchmark) vector, cross-validation and grouping
of variables (features, models, methods, etc.). The theory has been developed including the repeated
observations (ties):
(i) The exact theoretical distribution (null distribution) for 4 < number of objects < 9 has been calculated
for SRD treatment of random numbers (it's a kind of validation, a permutation test). All possible
reference vectors with ties imply different distribution.
(ii) For number of objects above eight (n > 8) an approximation has been developed using the Gaussian
distribution fitted on the SRD distribution given by generating of three million n-dimensional random
vectors.
The validity and features of the SRD methodology with ties are illustrated using two case studies: evaluation
of a sensory panel and ranking of financial indicators
Holmqvist, K., Nyström, N., Andersson, R., Dewhurst, R., Jarodzka, H., & Van de Weijer, J. (Eds.) (2011). Eye tracking: a comprehensive guide to methods and measures, Oxford, UK: Oxford University Press.
This review covers a novel approach to comparing methods, based on the sum of ranking differences (SRD). Many method-comparison studies suffer from ambiguity or from comparisons not being quite fair. This problem can be avoided if there are differences between ideal and actual rankings. The absolute values of differences for the ideal and actual ranking are summed up and the procedure is repeated for each (actual) method. The SRD values obtained such a way order the methods simply. If the ideal ranking is not known, it can be replaced by the average (maximum or minimum of all methods or by a known sequence).SRD corresponds to the principle of parsimony and provides an easy tool to evaluate the methods: the smaller the sum the better the method. Models and other items can be similarly ranked.Validation can be carried out using simulated random numbers for comparison: an empirical histogram (bootstrap-like) shows whether the SRD values are far from random.Two case studies (clustering of HPLC columns and prediction of retention data) illustrate and validate the applicability of this novel approach to comparing methods.The technique is entirely general; it can be used in different fields (e.g., for stationary-phase (column) selection in chromatography, model and descriptor selection, comparing analytical and chemometric techniques, determination of panel consistency, etc.). The only prerequisite is that the data can be arranged in matrix form without empty cells.
A training set of data has been used to construct a rule for predicting future responses. What is the error rate of this rule? The traditional answer to this question is given by cross-validation. The cross-validation estimate of prediction error is nearly unbiased, but can be highly variable. This article discusses bootstrap estimates of prediction error, which can be thought of as smoothed versions of cross-validation. A particular bootstrap method, the 632+ rule, is shown to substantially outperform cross-validation in a catalog of 24 simulation experiments. Besides providing point estimates, we also consider estimating the variability of an error rate estimate. All of the results here are nonparametric, and apply to any possible prediction rule: however we only study classification problems with 0-1 loss in detail. Our simulations include "smooth" prediction rules like Fisher's Linear Discriminant Function, and unsmooth ones like Nearest Neighbors. 1 Introduction This article conc...
Eye tracking the user experience
- A Bojko
Bojko, A. (2013). Eye tracking the user experience. Brooklyn, New York: Rosenfeld Media.
Eye tracking study recruitment - managing participants with vision irregularities
- Tobii
Tobii (2020). Eye tracking study recruitment -managing participants with vision
irregularities. Retrieved from https://www.tobiipro.com/blog/eye-tracking-study-re
cruitment-managing-participants-with-vision-irregularities/.