Dr. Ali Fallah Tehrani

Angewandte künstliche Intelligenz Maschinelles Lernen Datenanalyse

Wissenschaftlicher Mitarbeiter

Grafenau

08552/975620-40


JournalArticle
  • Ali Fallah-Tehrani
The Choquet Kernel on the use of Regression Problem
  • 2020

DOI: 10.1016/j.ins.2020.11.051

Recently, we have presented a new family of kernels on the basis of the discrete Choquet integral. While a naïve computation of this kernel has an exponential complexity in the number of features, we have proposed an efficient approach with computational complexity of 1. This kernel family is able to recognize dependencies between features and moreover it can be regularized through a proper selection of q-additivity. In fact, to reduce the effect of over-fitting there is an opportunity to restrict the flexibility of kernel to a lower degree. A key feature of the Choquet integral in a data-driven way is its monotonicity, however, this representation does not consider any monotonicity constraint; hence it is versatile for other applications, too. This issue is highlighted in the experimental study. In this regard, we apply the Choquet kernel for regression task and compare the performance of the proposed kernel versus state-of-the-art support kernel-based regression methods as well as random forest.
JournalArticle
  • Ali Fallah-Tehrani
  • M. Strickert
  • Diane Ahrens
Class of Monotone Kernelized Classifiers on the basis of the Choquet Integral , vol37
  • 2020

DOI: 10.1111/exsy.12506

The key property of monotone classifiers is that increasing (decreasing) input values lead to increasing (decreasing) the output value. Preserving monotonicity for a classifier typically requires many constraints to be respected by modeling approaches such as artificial intelligence techniques. The type of constraints strongly depends on the modeling assumptions. Of course, for sophisticated models such conditions might be very complex. In this study we present a new family of kernels that we call it Choquet kernels. Henceforth it allows for employing popular kernel‐based methods such as support vector machines. Instead of a naïve approach with exponential computational complexity we propose an equivalent formulation with quadratic time in the number of attributes. Furthermore, since coefficients derived from kernel solutions are not necessarily monotone in the dual form, different approaches are proposed to monotonize coefficients. Finally experiments illustrate beneficial properties of the Choquet kernels.
JournalArticle
  • Ali Fallah-Tehrani
The Choquet Kernel on the use of Regression Problem
  • 2020

DOI: 10.1016/j.ins.2020.11.051

Recently, we have presented a new family of kernels on the basis of the discrete Choquet integral. While a naïve computation of this kernel has an exponential complexity in the number of features, we have proposed an efficient approach with computational complexity of 1. This kernel family is able to recognize dependencies between features and moreover it can be regularized through a proper selection of q-additivity. In fact, to reduce the effect of over-fitting there is an opportunity to restrict the flexibility of kernel to a lower degree. A key feature of the Choquet integral in a data-driven way is its monotonicity, however, this representation does not consider any monotonicity constraint; hence it is versatile for other applications, too. This issue is highlighted in the experimental study. In this regard, we apply the Choquet kernel for regression task and compare the performance of the proposed kernel versus state-of-the-art support kernel-based regression methods as well as random forest.
Contribution
  • Monica I. Ciolacu
  • Ali Fallah-Tehrani
  • P. M. Svasta
  • I. Tache
  • D. Stoichescu
Education 4.0: An Adaptive Framework with Artificial Intelligence, Raspberry Pi and Wearables- Innovation for Creating Value
  • 2020

DOI: 10.1109/SIITME50350.2020.9292148

The Education 4.0 process can be used to foster students' performance, to motivate them by means of adaptive and personalized learning, to automate answering routine questions and to improve the quality of online examinations. It releases teachers from routine tasks, and it enables them to be more involved in the individualization of the educational process and in innovation. We apply artificial intelligence methodology to motivate students to improve their performance, and to identify early in course which students are at-risk of dropping out. In the adaptive Education 4.0 loT framework, the environmental and embedded sensors from wearables examine biosignals with biofeedback, such as pulse (HR) and heart rate variability (HRV), thereby increasing students' self-reflection, and allowing for a more personalized learning experience. The first experiments corroborate meaningful correlations between the data-points reached in the self-assessments, and values of the biosignals.
Contribution
  • Monica I. Ciolacu
  • Ali Fallah-Tehrani
  • P. M. Svasta
  • I. Tache
  • D. Stoichescu
Education 4.0: An Adaptive Framework with Artificial Intelligence, Raspberry Pi and Wearables- Innovation for Creating Value
  • 2020

DOI: 10.1109/SIITME50350.2020.9292148

The Education 4.0 process can be used to foster students' performance, to motivate them by means of adaptive and personalized learning, to automate answering routine questions and to improve the quality of online examinations. It releases teachers from routine tasks, and it enables them to be more involved in the individualization of the educational process and in innovation. We apply artificial intelligence methodology to motivate students to improve their performance, and to identify early in course which students are at-risk of dropping out. In the adaptive Education 4.0 loT framework, the environmental and embedded sensors from wearables examine biosignals with biofeedback, such as pulse (HR) and heart rate variability (HRV), thereby increasing students' self-reflection, and allowing for a more personalized learning experience. The first experiments corroborate meaningful correlations between the data-points reached in the self-assessments, and values of the biosignals.
JournalArticle
  • Ali Fallah-Tehrani
  • M. Strickert
  • Diane Ahrens
Class of Monotone Kernelized Classifiers on the basis of the Choquet Integral , vol37
  • 2020

DOI: 10.1111/exsy.12506

The key property of monotone classifiers is that increasing (decreasing) input values lead to increasing (decreasing) the output value. Preserving monotonicity for a classifier typically requires many constraints to be respected by modeling approaches such as artificial intelligence techniques. The type of constraints strongly depends on the modeling assumptions. Of course, for sophisticated models such conditions might be very complex. In this study we present a new family of kernels that we call it Choquet kernels. Henceforth it allows for employing popular kernel‐based methods such as support vector machines. Instead of a naïve approach with exponential computational complexity we propose an equivalent formulation with quadratic time in the number of attributes. Furthermore, since coefficients derived from kernel solutions are not necessarily monotone in the dual form, different approaches are proposed to monotonize coefficients. Finally experiments illustrate beneficial properties of the Choquet kernels.
Lecture
  • Ali Fallah-Tehrani
Echtzeit-Fehlererkennung und Verbesserung der Produktionsqualität in der Glasindustrie durch künstliche Intelligenz Posterpräsentation
  • 2019
JournalArticle
  • M. Aggarwal
  • Ali Fallah-Tehrani
Modelling Human Decision Behaviour with Preference Learning , vol31
  • 2019

DOI: 10.1287/ijoc.2018.0823

Preferences provide a means for specifying the desires of a decision maker (DM) in a declarative way. In this paper, based on a DM’s pairwise preferences, we infer the DM’s unique decision model. We capture (a) the attitudinal character, (b) relative criteria importance, and (c) the criteria interaction, all of which are specific to the DM. We make use of the preference-learning (PL) technique to induce predictive preference models from empirical data. Because PL is emerging as a new subfield of machine learning, we could use standard machine-learning methods to accomplish our learning objective. We consider the DM’s exemplary preference information in the form of pairwise comparisons between alternatives as the training information. The DM’s decision model is captured in terms of (a), (b), and (c), through the parameters of an attitudinal Choquet integral operator. The proposed learning approach is validated through an experimental study on 16 standard data sets. The superiority of the proposed method in terms of predictive accuracy and easier interpretability is shown both theoretically as well as empirically.
Lecture
  • Ali Fallah-Tehrani
Fehlererkennung in Echtzeit für Glasindustrie mit ML
  • 2019
JournalArticle
  • M. Aggarwal
  • Ali Fallah-Tehrani
Modelling Human Decision Behaviour with Preference Learning , vol31
  • 2019

DOI: 10.1287/ijoc.2018.0823

Preferences provide a means for specifying the desires of a decision maker (DM) in a declarative way. In this paper, based on a DM’s pairwise preferences, we infer the DM’s unique decision model. We capture (a) the attitudinal character, (b) relative criteria importance, and (c) the criteria interaction, all of which are specific to the DM. We make use of the preference-learning (PL) technique to induce predictive preference models from empirical data. Because PL is emerging as a new subfield of machine learning, we could use standard machine-learning methods to accomplish our learning objective. We consider the DM’s exemplary preference information in the form of pairwise comparisons between alternatives as the training information. The DM’s decision model is captured in terms of (a), (b), and (c), through the parameters of an attitudinal Choquet integral operator. The proposed learning approach is validated through an experimental study on 16 standard data sets. The superiority of the proposed method in terms of predictive accuracy and easier interpretability is shown both theoretically as well as empirically.
Lecture
  • Ali Fallah-Tehrani
Fehlererkennung in Echtzeit für Glasindustrie mit ML
  • 2019
Lecture
  • Ali Fallah-Tehrani
Echtzeit-Fehlererkennung und Verbesserung der Produktionsqualität in der Glasindustrie durch künstliche Intelligenz Posterpräsentation
  • 2019
Contribution
  • Monica I. Ciolacu
  • Ali Fallah-Tehrani
  • Leon Binder
  • P. M. Svasta
Education 4.0 - Artificial Intelligence Assisted Higher Education: Early Recognition System with Machine Learning to Support Students' Success
  • 2018

DOI: 10.1109/SIITME.2018.8599203

Education 4.0 is being empowered more and more by artificial intelligence (AI) methods. We observe a continuously growing demand for adaptive and personalized Education. In this paper we present an innovative approach to promoting AI in Education 4.0. Our first contribution is AI assisted Higher Education Process with smart sensors and wearable devices for self-regulated learning. Secondly we describe our first results of Education 4.0 didactic methods implemented with learning analytics and machine learning algorithms. The aim of this case study is to predict the final score of students before participating in final examination. We propose an Early Recognition System equipped with real data captured in a blended learning course with a personalized test at the beginning of the semester, an adaptive learning environment based on Auto Tutor by N. A. Crowder theory with adaptive self-assessment feedback. It is obvious that focusing on students' success and their experiences is a win-win scenario for students and professors as well as for the administration.
Contribution
  • Ali Fallah-Tehrani
  • Diane Ahrens
Enhanced Predictive Models for Purchasing in the Fashion Field by Applying Regression Trees Equipped with Ordinal Logistic Regression
  • 2018

DOI: 10.1007/978-981-13-0080-6_3

Identifying the products which are highly sold in the fashion apparel industry is one of the challenging tasks, which leads to reduce the write-off and increase the revenue. Assuming three classes as substantial, middle, and inconsiderable, the forecasting problem comes down to a classification problem, where the task is to predict the class of a product. In this research, we present a probabilistic approach to identify the class of fashion products in terms of sale. In previous work, we showed that a combination of kernel machines with a probabilistic approach may empower the performance of kernel machines. However, a well-known drawback of kernel machines is its non-interpretability. The interpretability is one of the most important features from an user point of view; essentially in the fashion field, decision makers require to understand and interpret the model for a more convenient adaptation. Since regression trees can be formulated through rules, this makes possible to comprehend the model. Nevertheless, a drawback of decision trees is the sensibility to input space, which may cause very enormous deviations in terms of prediction. To reduce this effect on forecast, we propose a new model equipped with ordinal logistic regression. Finally to verify the proposed approach, we conducted several experiments on a real data extracted from an apparel retailer in Germany.
Lecture
  • Monica I. Ciolacu
  • Ali Fallah-Tehrani
  • Leon Binder
  • P. Mugur Svasta
Education 4.0 - Artificial Intelligence Assisted Higher Education: Early Recognition System with Machine Learning to support Students' Success
  • 2018
Lecture
  • Monica I. Ciolacu
  • Ali Fallah-Tehrani
  • Leon Binder
  • P. Mugur Svasta
Education 4.0 - Artificial Intelligence Assisted Higher Education: Early Recognition System with Machine Learning to support Students' Success
  • 2018
Contribution
  • Monica I. Ciolacu
  • Ali Fallah-Tehrani
  • Leon Binder
  • P. M. Svasta
Education 4.0 - Artificial Intelligence Assisted Higher Education: Early Recognition System with Machine Learning to Support Students' Success
  • 2018

DOI: 10.1109/SIITME.2018.8599203

Education 4.0 is being empowered more and more by artificial intelligence (AI) methods. We observe a continuously growing demand for adaptive and personalized Education. In this paper we present an innovative approach to promoting AI in Education 4.0. Our first contribution is AI assisted Higher Education Process with smart sensors and wearable devices for self-regulated learning. Secondly we describe our first results of Education 4.0 didactic methods implemented with learning analytics and machine learning algorithms. The aim of this case study is to predict the final score of students before participating in final examination. We propose an Early Recognition System equipped with real data captured in a blended learning course with a personalized test at the beginning of the semester, an adaptive learning environment based on Auto Tutor by N. A. Crowder theory with adaptive self-assessment feedback. It is obvious that focusing on students' success and their experiences is a win-win scenario for students and professors as well as for the administration.
Contribution
  • Ali Fallah-Tehrani
  • Diane Ahrens
Enhanced Predictive Models for Purchasing in the Fashion Field by Applying Regression Trees Equipped with Ordinal Logistic Regression
  • 2018

DOI: 10.1007/978-981-13-0080-6_3

Identifying the products which are highly sold in the fashion apparel industry is one of the challenging tasks, which leads to reduce the write-off and increase the revenue. Assuming three classes as substantial, middle, and inconsiderable, the forecasting problem comes down to a classification problem, where the task is to predict the class of a product. In this research, we present a probabilistic approach to identify the class of fashion products in terms of sale. In previous work, we showed that a combination of kernel machines with a probabilistic approach may empower the performance of kernel machines. However, a well-known drawback of kernel machines is its non-interpretability. The interpretability is one of the most important features from an user point of view; essentially in the fashion field, decision makers require to understand and interpret the model for a more convenient adaptation. Since regression trees can be formulated through rules, this makes possible to comprehend the model. Nevertheless, a drawback of decision trees is the sensibility to input space, which may cause very enormous deviations in terms of prediction. To reduce this effect on forecast, we propose a new model equipped with ordinal logistic regression. Finally to verify the proposed approach, we conducted several experiments on a real data extracted from an apparel retailer in Germany.
Contribution
  • Monica I. Ciolacu
  • Ali Fallah-Tehrani
  • R. Beer
  • Heribert Popp
Education 4.0 – Fostering Student's Performance with Machine Learning Methods
  • 2017

DOI: 10.1109/SIITME.2017.8259941

JournalArticle
  • Ali Fallah-Tehrani
  • Diane Ahrens
Modeling Label Dependence for Multi-Label Classification Using the Choquistic Regression , vol92
  • 2017

DOI: 10.1016/j.patrec.2017.04.018

While an incorrect identification of underlying dependency in data can lead to a flawed conclusion, recognizing legitimate dependency allows for the opportunity to adapt a model in a correct manner. In this regard, modeling the inter-dependencies in multi-label classification (multi target prediction) is one of the challenging tasks from a machine learning point of view. While common approaches seek to exploit so-called correlated information from labels, this can be improved by assuming the interactions between labels. A well-known tool to model the interaction between attributes is the Choquet integral; it enables one to model non-linear dependencies between attributes. Beyond identifying proper prior knowledge in data (if such knowledge exists), establishing suitable models that are in agreement with prior knowledge is not always a trivial task. In this paper, we propose a first step towards modeling label dependencies for multi-target classifications in terms of positive and negative interactions. In the experimental, we demonstrate real gains by applying this approach.
JournalArticle
  • Ali Fallah-Tehrani
  • Diane Ahrens
Modified Sequential k‐means Clustering by Utilizing Response: A Case Study for Fashion Products , vol34
  • 2017

DOI: 10.1111/exsy.12226

Modified sequential k‐means clustering concerns a k‐means clustering problem in which the clustering machine utilizes output similarity in addition. While conventional clustering methods commonly recognize similar instances at features‐level modified sequential clustering takes advantage of response, too. To this end, the approach we pursue is to enhance the quality of clustering by using some proper information. The information enables the clustering machine to detect more patterns and dependencies that may be relevant. This allows one to determine, for instance, which fashion products exhibit similar behaviour in terms of sales. Unfortunately, conventional clustering methods cannot tackle such cases, because they handle attributes solely at the feature level without considering any response. In this study, we introduce a novel approach underlying minimum conditional entropy clustering and show its advantages in terms of data analytics. In particular, we achieve this by modifying the conventional sequential k‐means algorithm. This modified clustering approach has the ability to reflect the response effect in a consistent manner. To verify the feasibility and the performance of this approach, we conducted several experiments based on real data from the apparel industry.
JournalArticle
  • Ali Fallah-Tehrani
  • Diane Ahrens
Modeling Label Dependence for Multi-Label Classification Using the Choquistic Regression , vol92
  • 2017

DOI: 10.1016/j.patrec.2017.04.018

While an incorrect identification of underlying dependency in data can lead to a flawed conclusion, recognizing legitimate dependency allows for the opportunity to adapt a model in a correct manner. In this regard, modeling the inter-dependencies in multi-label classification (multi target prediction) is one of the challenging tasks from a machine learning point of view. While common approaches seek to exploit so-called correlated information from labels, this can be improved by assuming the interactions between labels. A well-known tool to model the interaction between attributes is the Choquet integral; it enables one to model non-linear dependencies between attributes. Beyond identifying proper prior knowledge in data (if such knowledge exists), establishing suitable models that are in agreement with prior knowledge is not always a trivial task. In this paper, we propose a first step towards modeling label dependencies for multi-target classifications in terms of positive and negative interactions. In the experimental, we demonstrate real gains by applying this approach.
Contribution
  • Monica I. Ciolacu
  • Ali Fallah-Tehrani
  • R. Beer
  • Heribert Popp
Education 4.0 – Fostering Student's Performance with Machine Learning Methods
  • 2017

DOI: 10.1109/SIITME.2017.8259941

JournalArticle
  • Ali Fallah-Tehrani
  • Diane Ahrens
Modified Sequential k‐means Clustering by Utilizing Response: A Case Study for Fashion Products , vol34
  • 2017

DOI: 10.1111/exsy.12226

Modified sequential k‐means clustering concerns a k‐means clustering problem in which the clustering machine utilizes output similarity in addition. While conventional clustering methods commonly recognize similar instances at features‐level modified sequential clustering takes advantage of response, too. To this end, the approach we pursue is to enhance the quality of clustering by using some proper information. The information enables the clustering machine to detect more patterns and dependencies that may be relevant. This allows one to determine, for instance, which fashion products exhibit similar behaviour in terms of sales. Unfortunately, conventional clustering methods cannot tackle such cases, because they handle attributes solely at the feature level without considering any response. In this study, we introduce a novel approach underlying minimum conditional entropy clustering and show its advantages in terms of data analytics. In particular, we achieve this by modifying the conventional sequential k‐means algorithm. This modified clustering approach has the ability to reflect the response effect in a consistent manner. To verify the feasibility and the performance of this approach, we conducted several experiments based on real data from the apparel industry.
JournalArticle
  • Ali Fallah-Tehrani
  • Diane Ahrens
Enhanced predictive models for purchasing in the fashion field by using kernel machine regression equipped with ordinal logistic regression , vol32
  • 2016
Identifying the products which are highly sold in the fashion apparel industry is one of the challenging tasks, which leads to reduce the write off and increases the revenue. In fact, beyond of sales forecasting in general a crucial question remains whether a product may sell well or not. Assuming three classes as substantial, middle and inconsiderable, the forecasting problem comes down to a classification problem, where the task is to predict the class of a product. In this research, we present a probabilistic approach to identify the class of fashion products in terms of sale. Thereafter, we combine kernel machines with a probabilistic approach to empower the performance of kernel machines and eventually to make use of it to predicting the number of sales. The proposed approach is more robust to outliers (in the case of highly sold products) and in addition uses prior knowledge, hence it serves more reliable results. In order to verify the proposed approach, we conducted several experiments on a real data extracted from an apparel retailer in Germany.
JournalArticle
  • Ali Fallah-Tehrani
  • Diane Ahrens
Supervised Regression Clustering: A Case Study for Fashion Products , vol3
  • 2016

DOI: 10.4018/IJBAN.2016100102

Clustering techniques typically group similar instances underlying individual attributes by supposing that similar instances have similar attributes characteristic. On contrary, clustering similar instances given a specific behavior is framed through supervised learning. For instance, which fashion products have similar behavior in term of sales. Unfortunately, conventional clustering methods cannot tackle this case, since they handle attributes by a same manner. In fact, conventional clustering approaches do not consider any response, and moreover they assume attributes act by the same importance. However, clustering instances with respect to responses leads to a better data analytics. In this research, the authors introduce an approach for the goal supervised clustering and show its advantage in terms of data analytics as well as prediction. To verify the feasibility and the performance of this approach the authors conducted several experiments on a real dataset derived from an apparel industry.
Contribution
  • Ali Fallah-Tehrani
  • Diane Ahrens
Improved Forecasting and Purchasing of Fashion Products based on the Use of Big Data Techniques
  • 2016

DOI: 10.1007/978-3-658-08809-5_13

Ordering proper amount of products, taking into account the demand of market, in fashion retail industries is one of the core challenges. Essentially due to the fact that the ordering is typically performed once in the each season, it is absolutely required to carry out precise orders. To make a precise ordering as well as to prevent overstocks and stock-out, there is a need for reliable forecasting methods. A reliable forecasting requires to consider proper predictive models which can consider all deciding factors. Specifically in the case of fashion forecasting since each product is associated with several factors, e.g. price, style, color and even human factors, learn a suitable predictive model is not an easy task. In fact, the challenge here boils down to learn a powerful model, which can cover all these information. To this end, big data techniques, namely data mining and machine learning methods serve the ability to accomplish the challenge. In this paper, we exploit unsupervised learning methods for a goal fitting the data, particularly w.r.t simple models although with higher gain. In essence, our innovative model is able to modify simple regression model, and hence, provide more promising results. In this regard, we apply big data analyses and techniques specifically in fashion field to analyze and make the salesprediction.
Lecture
  • Ali Fallah-Tehrani
Was Frauen wollen - Absatzprognosen im Modehandel durch künstliche Intelligenz
  • 2016
Contribution
  • Ali Fallah-Tehrani
  • Diane Ahrens
Improved Forecasting and Purchasing of Fashion Products based on the Use of Big Data Techniques
  • 2016

DOI: 10.1007/978-3-658-08809-5_13

Ordering proper amount of products, taking into account the demand of market, in fashion retail industries is one of the core challenges. Essentially due to the fact that the ordering is typically performed once in the each season, it is absolutely required to carry out precise orders. To make a precise ordering as well as to prevent overstocks and stock-out, there is a need for reliable forecasting methods. A reliable forecasting requires to consider proper predictive models which can consider all deciding factors. Specifically in the case of fashion forecasting since each product is associated with several factors, e.g. price, style, color and even human factors, learn a suitable predictive model is not an easy task. In fact, the challenge here boils down to learn a powerful model, which can cover all these information. To this end, big data techniques, namely data mining and machine learning methods serve the ability to accomplish the challenge. In this paper, we exploit unsupervised learning methods for a goal fitting the data, particularly w.r.t simple models although with higher gain. In essence, our innovative model is able to modify simple regression model, and hence, provide more promising results. In this regard, we apply big data analyses and techniques specifically in fashion field to analyze and make the salesprediction.
JournalArticle
  • Ali Fallah-Tehrani
  • Diane Ahrens
Supervised Regression Clustering: A Case Study for Fashion Products , vol3
  • 2016

DOI: 10.4018/IJBAN.2016100102

Clustering techniques typically group similar instances underlying individual attributes by supposing that similar instances have similar attributes characteristic. On contrary, clustering similar instances given a specific behavior is framed through supervised learning. For instance, which fashion products have similar behavior in term of sales. Unfortunately, conventional clustering methods cannot tackle this case, since they handle attributes by a same manner. In fact, conventional clustering approaches do not consider any response, and moreover they assume attributes act by the same importance. However, clustering instances with respect to responses leads to a better data analytics. In this research, the authors introduce an approach for the goal supervised clustering and show its advantage in terms of data analytics as well as prediction. To verify the feasibility and the performance of this approach the authors conducted several experiments on a real dataset derived from an apparel industry.
JournalArticle
  • Ali Fallah-Tehrani
  • Diane Ahrens
Enhanced predictive models for purchasing in the fashion field by using kernel machine regression equipped with ordinal logistic regression , vol32
  • 2016
Identifying the products which are highly sold in the fashion apparel industry is one of the challenging tasks, which leads to reduce the write off and increases the revenue. In fact, beyond of sales forecasting in general a crucial question remains whether a product may sell well or not. Assuming three classes as substantial, middle and inconsiderable, the forecasting problem comes down to a classification problem, where the task is to predict the class of a product. In this research, we present a probabilistic approach to identify the class of fashion products in terms of sale. Thereafter, we combine kernel machines with a probabilistic approach to empower the performance of kernel machines and eventually to make use of it to predicting the number of sales. The proposed approach is more robust to outliers (in the case of highly sold products) and in addition uses prior knowledge, hence it serves more reliable results. In order to verify the proposed approach, we conducted several experiments on a real data extracted from an apparel retailer in Germany.
Lecture
  • Ali Fallah-Tehrani
Was Frauen wollen - Absatzprognosen im Modehandel durch künstliche Intelligenz
  • 2016
Lecture
  • Ali Fallah-Tehrani
Learning Classifiers on the Use of Monotone Learning
  • 2015
Lecture
  • Ali Fallah-Tehrani
Learning Classifiers on the Use of Monotone Learning
  • 2015
Contribution
  • Ali Fallah-Tehrani
  • M. Strickert
  • E. Hüllermeier
The Choquet Kernel for Monotone Data
  • 2014
Contribution
  • Ali Fallah-Tehrani
  • C. Lebreuche
  • E. Hüllermeier
Utilitaristic Choquistic Regression
  • 2014
JournalArticle
  • M. Agarwal
  • Ali Fallah-Tehrani
  • E. Hüllermeier
Preference-based Learning of Ideal Solutions in TOPSIS-like Decision Models , vol22
  • 2014

DOI: 10.1002/mcda.1520

Combining established modelling techniques from multiple-criteria decision aiding with recent algorithmic advances in the emerging field of preference learning, we propose a new method that can be seen as an adaptive version of TOPSIS, the technique for order preference by similarity to ideal solution decision model (or at least a simplified variant of this model). On the basis of exemplary preference information in the form of pairwise comparisons between alternatives, our method seeks to induce an ‘ideal solution’ that, in conjunction with a weight factor for each criterion, represents the preferences of the decision maker. To this end, we resort to probabilistic models of discrete choice and make use of maximum likelihood inference. First experimental results on suitable preference data suggest that our approach is not only intuitively appealing and interesting from an interpretation point of view but also competitive to state-of-the-art preference learning methods in terms of prediction accuracy.
Contribution
  • Ali Fallah-Tehrani
  • C. Lebreuche
  • E. Hüllermeier
Utilitaristic Choquistic Regression
  • 2014
JournalArticle
  • M. Agarwal
  • Ali Fallah-Tehrani
  • E. Hüllermeier
Preference-based Learning of Ideal Solutions in TOPSIS-like Decision Models , vol22
  • 2014

DOI: 10.1002/mcda.1520

Combining established modelling techniques from multiple-criteria decision aiding with recent algorithmic advances in the emerging field of preference learning, we propose a new method that can be seen as an adaptive version of TOPSIS, the technique for order preference by similarity to ideal solution decision model (or at least a simplified variant of this model). On the basis of exemplary preference information in the form of pairwise comparisons between alternatives, our method seeks to induce an ‘ideal solution’ that, in conjunction with a weight factor for each criterion, represents the preferences of the decision maker. To this end, we resort to probabilistic models of discrete choice and make use of maximum likelihood inference. First experimental results on suitable preference data suggest that our approach is not only intuitively appealing and interesting from an interpretation point of view but also competitive to state-of-the-art preference learning methods in terms of prediction accuracy.
Contribution
  • Ali Fallah-Tehrani
  • M. Strickert
  • E. Hüllermeier
The Choquet Kernel for Monotone Data
  • 2014
Contribution
  • Ali Fallah-Tehrani
  • E. Hüllermeier
Ordinal Choquistic Regression
  • 2013
Contribution
  • E. Hüllermeier
  • Ali Fallah-Tehrani
Efficient Learning of Classifiers based on the 2-additive Choquet Integral Computational Intelligence
  • 2013
In a recent work, we proposed a generalization of logistic regression based on the Choquet integral. Our approach, referred to as choquistic regression, makes it possible to capture non-linear dependencies and interactions among predictor variables while preserving two important properties of logistic regression, namely the comprehensibility of the model and the possibility to ensure its monotonicity in individual predictors. Unsurprisingly, these benefits come at the expense of an increased computational complexity of the underlying maximum likelihood estimation. In this paper, we propose two approaches for reducing this complexity in the specific though practically relevant case of the 2-additive Choquet integral. Apart from theoretical results, we also present an experimental study in which we compare the two variants with the original implementation of choquistic regression.
Contribution
  • E. Hüllermeier
  • Ali Fallah-Tehrani
Efficient Learning of Classifiers based on the 2-additive Choquet Integral Computational Intelligence
  • 2013
In a recent work, we proposed a generalization of logistic regression based on the Choquet integral. Our approach, referred to as choquistic regression, makes it possible to capture non-linear dependencies and interactions among predictor variables while preserving two important properties of logistic regression, namely the comprehensibility of the model and the possibility to ensure its monotonicity in individual predictors. Unsurprisingly, these benefits come at the expense of an increased computational complexity of the underlying maximum likelihood estimation. In this paper, we propose two approaches for reducing this complexity in the specific though practically relevant case of the 2-additive Choquet integral. Apart from theoretical results, we also present an experimental study in which we compare the two variants with the original implementation of choquistic regression.
Contribution
  • Ali Fallah-Tehrani
  • E. Hüllermeier
Ordinal Choquistic Regression
  • 2013
JournalArticle
  • Ali Fallah-Tehrani
  • W. Cheng
  • K. Dembczýnski
  • E. Hüllermeier
Learning Monotone Nonlinear Models using the Choquet Integral , vol89
  • 2012

DOI: 10.1007/s10994-012-5318-3

The learning of predictive models that guarantee monotonicity in the input variables has received increasing attention in machine learning in recent years. By trend, the difficulty of ensuring monotonicity increases with the flexibility or, say, nonlinearity of a model. In this paper, we advocate the so-called Choquet integral as a tool for learning monotone nonlinear models. While being widely used as a flexible aggregation operator in different fields, such as multiple criteria decision making, the Choquet integral is much less known in machine learning so far. Apart from combining monotonicity and flexibility in a mathematically sound and elegant manner, the Choquet integral has additional features making it attractive from a machine learning point of view. Notably, it offers measures for quantifying the importance of individual predictor variables and the interaction between groups of variables. Analyzing the Choquet integral from a classification perspective, we provide upper and lower bounds on its VC-dimension. Moreover, as a methodological contribution, we propose a generalization of logistic regression. The basic idea of our approach, referred to as choquistic regression, is to replace the linear function of predictor variables, which is commonly used in logistic regression to model the log odds of the positive class, by the Choquet integral. First experimental results are quite promising and suggest that the combination of monotonicity and flexibility offered by the Choquet integral facilitates strong performance in practical applications.
JournalArticle
  • Ali Fallah-Tehrani
  • W. Cheng
  • E. Hüllermeier
Preference Learning using the Choquet Integral: The Case of Multipartite Ranking , vol20
  • 2012

DOI: 10.1109/TFUZZ.2012.2196050

We propose a novel method for preference learning or, more specifically, learning to rank, where the task is to learn a ranking model that takes a subset of alternatives as input and produces a ranking of these alternatives as output. Just like in the case of conventional classifier learning, training information is provided in the form of a set of labeled instances, with labels or, say, preference degrees taken from an ordered categorical scale. This setting is known as multipartite ranking in the literature. Our approach is based on the idea of using the (discrete) Choquet integral as an underlying model for representing ranking functions. Being an established aggregation function in fields such as multiple criteria decision making and information fusion, the Choquet integral offers a number of interesting properties that make it attractive from a machine learning perspective, too. The learning problem itself comes down to properly specifying the fuzzy measure on which the Choquet integral is defined. This problem is formalized as a margin maximization problem and solved by means of a cutting plane algorithm. The performance of our method is tested on a number of benchmark datasets.
JournalArticle
  • Ali Fallah-Tehrani
  • W. Cheng
  • K. Dembczýnski
  • E. Hüllermeier
Learning Monotone Nonlinear Models using the Choquet Integral , vol89
  • 2012

DOI: 10.1007/s10994-012-5318-3

The learning of predictive models that guarantee monotonicity in the input variables has received increasing attention in machine learning in recent years. By trend, the difficulty of ensuring monotonicity increases with the flexibility or, say, nonlinearity of a model. In this paper, we advocate the so-called Choquet integral as a tool for learning monotone nonlinear models. While being widely used as a flexible aggregation operator in different fields, such as multiple criteria decision making, the Choquet integral is much less known in machine learning so far. Apart from combining monotonicity and flexibility in a mathematically sound and elegant manner, the Choquet integral has additional features making it attractive from a machine learning point of view. Notably, it offers measures for quantifying the importance of individual predictor variables and the interaction between groups of variables. Analyzing the Choquet integral from a classification perspective, we provide upper and lower bounds on its VC-dimension. Moreover, as a methodological contribution, we propose a generalization of logistic regression. The basic idea of our approach, referred to as choquistic regression, is to replace the linear function of predictor variables, which is commonly used in logistic regression to model the log odds of the positive class, by the Choquet integral. First experimental results are quite promising and suggest that the combination of monotonicity and flexibility offered by the Choquet integral facilitates strong performance in practical applications.
JournalArticle
  • Ali Fallah-Tehrani
  • W. Cheng
  • E. Hüllermeier
Preference Learning using the Choquet Integral: The Case of Multipartite Ranking , vol20
  • 2012

DOI: 10.1109/TFUZZ.2012.2196050

We propose a novel method for preference learning or, more specifically, learning to rank, where the task is to learn a ranking model that takes a subset of alternatives as input and produces a ranking of these alternatives as output. Just like in the case of conventional classifier learning, training information is provided in the form of a set of labeled instances, with labels or, say, preference degrees taken from an ordered categorical scale. This setting is known as multipartite ranking in the literature. Our approach is based on the idea of using the (discrete) Choquet integral as an underlying model for representing ranking functions. Being an established aggregation function in fields such as multiple criteria decision making and information fusion, the Choquet integral offers a number of interesting properties that make it attractive from a machine learning perspective, too. The learning problem itself comes down to properly specifying the fuzzy measure on which the Choquet integral is defined. This problem is formalized as a margin maximization problem and solved by means of a cutting plane algorithm. The performance of our method is tested on a number of benchmark datasets.
Contribution
  • Ali Fallah-Tehrani
  • W. Cheng
  • K. Dembczýnski
  • E. Hüllermeier
Learning Monotone Nonlinear Models using the Choquet Integral
  • 2011
Contribution
  • Ali Fallah-Tehrani
  • W. Cheng
  • K. Dembczýnski
  • E. Hüllermeier
Learning Monotone Nonlinear Models using the Choquet Integral
  • 2011
Contribution
  • E. Hüllermeier
  • Ali Fallah-Tehrani
On the VC-Dimension of the Choquet Integral
Contribution
  • E. Hüllermeier
  • Ali Fallah-Tehrani
On the VC-Dimension of the Choquet Integral