scholarly journals Knowledge Tracing Machines: Factorization Machines for Knowledge Tracing

Author(s):  
Jill-Jenn Vie ◽  
Hisashi Kashima

Knowledge tracing is a sequence prediction problem where the goal is to predict the outcomes of students over questions as they are interacting with a learning platform. By tracking the evolution of the knowledge of some student, one can optimize instruction. Existing methods are either based on temporal latent variable models, or factor analysis with temporal features. We here show that factorization machines (FMs), a model for regression or classification, encompasses several existing models in the educational literature as special cases, notably additive factor model, performance factor model, and multidimensional item response theory. We show, using several real datasets of tens of thousands of users and items, that FMs can estimate student knowledge accurately and fast even when student data is sparsely observed, and handle side information such as multiple knowledge components and number of attempts at item or skill level. Our approach allows to fit student models of higher dimension than existing models, and provides a testbed to try new combinations of features in order to improve existing models.

2002 ◽  
Vol 29 (2) ◽  
pp. 161-182 ◽  
Author(s):  
Lening Zhang ◽  
John W. Welte ◽  
William F. Wieczorek

The Buffalo Longitudinal Study of Young Men was used to address the possibility of a common factor underlying adolescent problem behaviors. First, a measurement model with a single first-order factor was compared to a model with three separate correlated first-order factors. The three-factor model was better supported, making it logical to conduct a second-order factor analysis, which confirmed the logic. Second, a substantive model was estimated in each of two waves with psychopathic state as the common factor predicting drinking, drug use, and delinquency. Psychopathic state was stable across waves. The theory that a single latent variable accounts for large covariance among adolescent problem behaviors was supported.


Mathematics ◽  
2022 ◽  
Vol 10 (1) ◽  
pp. 142
Author(s):  
Konstantin B. Kostin ◽  
Philippe Runge ◽  
Michel Charifzadeh

This study empirically analyzes and compares return data from developed and emerging market data based on the Fama French five-factor model and compares it to previous results from the Fama French three-factor model by Kostin, Runge and Adams (2021). It researches whether the addition of the profitability and investment pattern factors show superior results in the assessment of emerging markets during the COVID-19 pandemic compared to developed markets. We use panel data covering eight indices of developed and emerging countries as well as a selection of eight companies from these markets, covering a period from 2000 to 2020. Our findings suggest that emerging markets do not generally outperform developed markets. The results underscore the need to reconsider the assumption that adding more factors to regression models automatically yields results that are more reliable. Our study contributes to the extant literature by broadening this research area. It is the first study to compare the performance of the Fama French three-factor model and the Fama French five-factor model in the cost of equity calculation for developed and emerging countries during the COVID-19 pandemic and other crisis events of the past two decades.


Author(s):  
Yinan Yang ◽  
Yingying Meng

Health is the key to the aging problem, and “healthy aging” depicts the overall changing trends in the health of all elderly individuals in a society. Based on the Chinese Longitudinal Healthy Longevity Survey (CLHLS) data from the years 2002, 2005, 2008, 2011 and 2014, this article investigates whether there is a trend of “healthy aging” in China. A second-order factor model including four dimensions of physical health, functional status, mental health and social health was constructed to measure a latent variable, “Health_elders”. The further multigroup comparison results by structural equation modeling showed that, with the exception of 2008, the Health_elders in 2002, 2005, 2011 and 2014 displayed an upward trend, and the mean differences in Health_elders across five periods were significant. These findings indicate that on the whole, compared with older people in the past, older people in more recent periods are healthier, which supports the trend of “healthy aging” in China. In terms of cohorts, the average health levels of male, town-residing elderly populations are higher, while the healthy aging trends among female, rural and urban elderly populations are stronger. Moreover, the physical health levels of the 60–74 years-old cohort are decreasing, and the participation of elderly individuals in social activities is low, which are the weaknesses in the healthy aging process in China.


Author(s):  
Zhiwei Wang ◽  
Xiaoqin Feng ◽  
Jiliang Tang ◽  
Gale Yan Huang ◽  
Zitao Liu

2016 ◽  
Vol 24 (1) ◽  
pp. 97-118
Author(s):  
Da Hea Kim ◽  
Tai-Yong Roh ◽  
Suk Joon Byun ◽  
Jung Soon Hyun

This study examines the empirical performance of emission allowance option pricing models, concentrating on the EU-ETS markets. For option pricing, we use parameters estimated from option market data whereas few papers in the extant literature use parameters from underlying asset data. As results, it is shown that the most appropriate is the one-factor model in which the EUA logarithmic spot prices follow the mean-reverting process of Ornstein-Uhlenbeck type. Also, the addition of jumps is not shown to make any significant improvement in the model performance. These results are quite striking in the sense that existing papers report the addition of jumps is necessary to improve option pricing on the EU-ETS markets. In sum, this paper is meaningful in the sense that it extends the growing empirical literature on the behavior of emission allowance spot prices


2009 ◽  
Vol 105 (2) ◽  
pp. 411-426 ◽  
Author(s):  
Denise Jepsen ◽  
John Rodwell

Dimensionality of the Colquitt justice measures was investigated across a wide range of service occupations. Structural equation modeling of data from 410 survey respondents found support for the 4-factor model of justice (procedural, distributive, interpersonal, and informational), although significant improvement of model fit was obtained by including a new latent variable, “procedural voice,” which taps employees' desire to express their views and feelings and influence results. The model was confirmed in a second sample ( N = 505) in the same organization six months later.


2005 ◽  
Vol 30 (1) ◽  
pp. 27-58 ◽  
Author(s):  
Bengt Muthén ◽  
Katherine Masyn

This article proposes a general latent variable approach to discrete-time survival analysis of nonrepeatable events such as onset of drug use. It is shown how the survival analysis can be formulated as a generalized latent class analysis of event history indicators. The latent class analysis can use covariates and can be combined with the joint modeling of other outcomes such as repeated measures for a related process. It is shown that conventional discrete-time survival analysis corresponds to a single-class latent class analysis. Multiple-class extensions are proposed, including the special cases of a class of long-term survivors and classes defined by outcomes related to survival. The estimation uses a general latent variable framework, including both categorical and continuous latent variables and incorporated in the Mplus program. Estimation is carried out using maximum likelihood via the EM algorithm. Two examples serve as illustrations. The first example concerns recidivism after incarceration in a randomized field experiment. The second example concerns school removal related to the development of aggressive behavior in the classroom.


2015 ◽  
Vol 734 ◽  
pp. 495-498
Author(s):  
Qing Feng Li

There are some bottleneck problems in the supervised machine learning and unsupervised machine learning. In view of the current problems, this paper tries to make some meaningful exploration. The main work is as follows: Research on the statistical analysis of factor analysis and latent variable and in some valuable research results of typical machine learning, and some no analysis method and factor analysis of supervised learning or hidden variables method to contact with the typical analysis, summary of the comprehensive characteristics of implicit factor model and to reveal the hiding data structures help and contributions.


Entropy ◽  
2021 ◽  
Vol 23 (3) ◽  
pp. 265
Author(s):  
Ran Tamir (Averbuch) ◽  
Neri Merhav

Typical random codes (TRCs) in a communication scenario of source coding with side information in the decoder is the main subject of this work. We study the semi-deterministic code ensemble, which is a certain variant of the ordinary random binning code ensemble. In this code ensemble, the relatively small type classes of the source are deterministically partitioned into the available bins in a one-to-one manner. As a consequence, the error probability decreases dramatically. The random binning error exponent and the error exponent of the TRCs are derived and proved to be equal to one another in a few important special cases. We show that the performance under optimal decoding can be attained also by certain universal decoders, e.g., the stochastic likelihood decoder with an empirical entropy metric. Moreover, we discuss the trade-offs between the error exponent and the excess-rate exponent for the typical random semi-deterministic code and characterize its optimal rate function. We show that for any pair of correlated information sources, both error and excess-rate probabilities exponential vanish when the blocklength tends to infinity.


2017 ◽  
Author(s):  
Yi Mou ◽  
Bo Zhang ◽  
Manuela Piazza ◽  
Daniel C. Hyde

Children’s understanding of the cardinal numbers before entering school provides the foundation for formal mathematics learning. Two types of tasks have been primarily used to measure children’s knowledge of cardinal numbers: set-to-number and number-to-set tasks. However, there has been a continued debate as to whether the two types of tasks measure the same conceptual construct, allowing comparison and interchangeable use, or whether they measure different but related constructs. To answer this question, we analyzed the relation between task and item level performance on representative set-to-number (e.g., How-Many?) and number-to-set (Give-N) tasks in a large group of 3- to 4-year-old preschoolers (N = 204, median age = 3y 10m). By constructing and comparing models with different latent variable structures, we found that the best-fitting model was a bi-factor model, where performance on set-to-number and number-to-set tasks is best explained by both overlapping and some distinct aspects of cardinal number knowledge. Further analyses ruled out the idea that differences between tasks were due solely to non-numerical, general cognitive or language factors. Together these results suggest that set-to-number and number-to-set tasks have some commonalities but also retain at least some significant conceptual distinctness. Based on these results, we suggest these two types of tasks should no longer be used indiscriminately to inform theory or educational assessment of numerical abilities in preschool children.


Sign in / Sign up

Export Citation Format

Share Document