KR20200137655A - Method and apparatus of prediction model for learning outcomes with perceived affordances in video-based learning environment - Google Patents

Abstract

Disclosed are a method for deriving a learning outcome prediction model through perceived affordance in a video-based learning environment, and an apparatus thereof. According to the present invention, a method for driving a learning output prediction model may comprise the steps of: collecting behavioral data and physiological and psychological data of a learner generated in the learning process for learning content; determining variables for the behavioral data and physiological and psychological data of the learner collected in the learning process through preprocessing of the collected behavioral data and physiological and psychological data of the learner; and deriving learning outcome prediction models for the entire group and detailed groups of the entire group based on the determined variables for the behavioral data and physiological and psychological data of the learner.

Description

A method and apparatus for deriving a model for predicting learning outcomes through perceived affordance in a video-based learning environment {METHOD AND APPARATUS OF PREDICTION MODEL FOR LEARNING OUTCOMES WITH PERCEIVED AFFORDANCES IN VIDEO-BASED LEARNING ENVIRONMENT}

The present invention relates to a method and apparatus for deriving a learning outcome prediction model, and more specifically, the present invention uses Affordance, a behavioral index that can more objectively represent the interaction between learners and media occurring in a video learning environment. It relates to the technology that predicts learners' learning outcomes.

The development of Information and Communication Technology (ICT) has led to the growth of various digital-based learning such as e-Learning, online learning, and web-based learning. Recently, it is spreading more rapidly with the advent of new video-based learning environments such as Massive Open Online Courses (MOOC), Khan Academy, TED and YouTube. Accordingly, there are increasing attempts at home and abroad to propose more improved teaching-learning designs with a deeper understanding of video-based learning.

The learning experienced by learners in a video-based learning environment is different from learning in a traditional face-to-face learning environment in that learners and instructors are separated. In a video-based learning environment, passive learning is achieved through one-way learning content delivery, and it is difficult to provide timely instructional treatment through real-time interaction between instructor and learner.

In video-based learning, learners' self-regulation ability is emphasized in that not only the space and time in which learning takes place, but also a series of learning processes depend on the learner's learning will. For successful learning in a video learning environment, it is important to promote learners to actively and self-directed learning. Accordingly, it is revealed that in order to perform effective learning in a video-based learning environment, a different teaching-learning strategy and model must be applied to that in a face-to-face learning environment.

In other words, designing a strategy that promotes learners' self-directed learning through interaction with a video-based learning environment is meaningful in improving learning outcomes. To do this, we can consider'perceived affordances'. Affordance refers to the nature of the environment that influences the behavior of the actor when the actor and the environment interact.

In order to understand the perceived affordance in a video-based learning environment, it is necessary to understand the inner process of the learner experienced in the process of interaction with the learning medium. Until now, most studies related to learning have measured learners' internal processes mainly through self-report questionnaires. In a video-based learning environment where learners and instructors are completely separated, instructors cannot observe the learning process, so learner behaviors that occur during learning activities cannot be grasped in detail. In addition, it is possible to know about learners' overall subjective thoughts and feelings about learning through a post-mortem self-report questionnaire, but it is difficult to obtain specific information about what experiences the learners have in which learning scenes.

Therefore, in order to adaptively cope with the changed learning environment, there is a need for research on the use of new data that can represent and measure the inner process of learners that change according to the video-based learning process.

The present invention provides a method and apparatus for collecting behavioral data and physiological psychological data representing the learner's perceptual affordance generated in the learning process for learning content, and deriving a learning outcome prediction model for each group using the collected data. I can.

In addition, the present invention can provide a method and apparatus for providing a required teaching-learning prescription to learners based on the derived model for predicting learning outcomes for each group.

An apparatus for deriving a learning outcome prediction model according to an embodiment of the present invention includes: collecting behavioral data and physiological psychological data of a learner generated in a learning process for learning content; Determining variables for the learner's behavioral data and physiological psychological data collected in the learning process through preprocessing of the collected learner's behavioral data and physiological psychological data; And deriving models for predicting learning outcomes for the entire group and detailed groups of the entire group based on the determined learner's behavioral data and variables for physiological and psychological data.

The deriving may include performing descriptive statistical analysis for the entire group based on the determined learner's behavioral data and variables for physiological and psychological data; And deriving a learning outcome prediction model for the entire group on which the descriptive statistical analysis was performed by using the random forest technique.

The deriving may include dividing the entire group into a first subgroup according to a task complexity level by performing a difference test based on the determined learner's behavior data and variables for physiological and psychological data; And deriving a learning outcome prediction model for the first subgroup by using the random forest technique.

The deriving may include performing cluster analysis on the divided first sub-groups and dividing the first sub-group into second sub-groups according to characteristics of each group, behavioral data, and physiological psychological data; And deriving a learning outcome prediction model for the second subgroup by using the random forest technique.

It may include the step of providing a teaching-learning prescription to the learner through comparison of the derived learning outcome prediction models by changing patterns and important variables.

The apparatus for deriving a learning outcome prediction model according to an embodiment of the present invention includes a processor, wherein the processor collects behavioral data and physiological psychological data of a learner generated in a learning process for learning content, and the collected learner's behavior The learner's behavioral data and variables for physiological psychological data for the learning process are determined through pre-processing of data and physiological psychological data, and based on the determined variables for the learner's behavioral data and physiological psychological data It is possible to derive a model for predicting learning outcomes for detailed groups of the entire group.

The processor may perform descriptive statistical analysis on the entire group based on the determined variables for the learner's behavioral data and physiological psychological data, and derive a learning outcome prediction model for the entire group using a random forest technique. have.

The processor divides the entire group into a first subgroup according to a task complexity level by performing a difference test based on the determined variables of the learner's behavioral data and physiological psychological data, and uses the random forest technique to divide the entire group into a first subgroup. 1 A model for predicting learning outcomes for subgroups can be derived.

The processor performs cluster analysis on the divided first sub-groups, divides the first sub-group into second sub-groups according to characteristics of each group, behavioral data, and physiological psychological data, and uses the random forest technique. Using this, a model for predicting learning outcomes for the second subgroup can be derived.

The processor may provide a teaching-learning prescription to the learner through a comparison of a change pattern of the derived learning outcome prediction model and each important variable.

The present invention collects behavioral data and physiological psychological data representing a learner's perceptual affordance generated in a learning process for learning content, and can derive a learning outcome prediction model for each group using the collected data.

In addition, the present invention can provide a teaching-learning prescription necessary for learners based on the derived model for predicting learning outcomes for each group.

1 is a diagram illustrating a system for deriving a learning outcome prediction model performed by an apparatus for deriving a learning outcome prediction model according to an embodiment of the present invention.
2 is a flowchart showing a method of deriving a learning outcome prediction model according to an embodiment of the present invention.
3 is a diagram illustrating an example of an AOI area activation screen according to an embodiment of the present invention.
4 is a flowchart illustrating a detailed method of deriving models for predicting learning outcomes for the entire group and detailed groups of the entire group according to an embodiment of the present invention.

Hereinafter, exemplary embodiments will be described in detail with reference to the accompanying drawings. However, since various changes may be made to the embodiments, the scope of the rights of the patent application is not limited or limited by these embodiments. It should be understood that all changes, equivalents, or substitutes to the embodiments are included in the scope of the rights.

The terms used in the examples are used for illustrative purposes only and should not be interpreted as limiting. Singular expressions include plural expressions unless the context clearly indicates otherwise. In the present specification, terms such as "comprise" or "have" are intended to designate the presence of features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, but one or more other features. It is to be understood that the presence or addition of elements or numbers, steps, actions, components, parts, or combinations thereof, does not preclude in advance.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiment belongs. Terms as defined in a commonly used dictionary should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and should not be interpreted as an ideal or excessively formal meaning unless explicitly defined in this application. Does not.

In addition, in the description with reference to the accompanying drawings, the same components are assigned the same reference numerals regardless of the reference numerals, and redundant descriptions thereof will be omitted. In describing the embodiments, when it is determined that a detailed description of related known technologies may unnecessarily obscure the subject matter of the embodiments, the detailed description thereof will be omitted.

1 is a diagram illustrating a system for deriving a learning outcome prediction model performed by an apparatus for deriving a learning outcome prediction model according to an embodiment of the present invention.

1, the learning outcome prediction model derivation apparatus 100 of the present invention collects learner's behavioral data and physiological psychological data in a video-based learning environment, and learns using the collected learner's behavioral data and physiological psychological data. You can derive a performance prediction model.

At this time, the learning outcome prediction model derivation device 100 verifies the effect of the perceptual affordance represented by learner's behavior data (ex. behavior log data) and physiological psychological data (ex. eye movement data) on learning outcomes in multiple ways. And, it is possible to derive a model for predicting learning outcomes according to the overall group, the first subgroup according to the task complexity level, and the second subgroup according to the cluster characteristics of the first subgroup.

Affordance is the possibility of inducing behavior that induces learners' actions, and is a functional characteristic of a technical environment that supports learners' goals and achievements. Learners take action only in the context of their purpose, and the action itself does not contain any specific meaning. However, affordances through learners' perceptions can meaningfully encompass learners' goals and motivations by allowing learners to interact with the environment.

In order to empirically extract the perceived affordance, the learner's behavior interacting with the domain of the technical environment can be considered. Affordance is essentially functional, and perceived affordance that directly affects the process of interaction with the environment must be considered not only in the social context in which the behavior occurs, but also in the learner's perspective.

Behavior log data, a kind of behavior data, is a representative variable that can extract perceived affordance. Such behavior log data is data that can be generated by learners in a video-based learning environment. Various types of data such as play, pause, search/move view point, bookmark registration/check, comment registration/check, annotation filtering, slide movement, playback environment adjustment, etc. Can contain kinds of data.

At this time, the behavior log data for a specific behavior may imply the behavior inducement for the environment related to the learner. In the context of learning, learner behavior log data in a learning environment can be collected and analyzed to identify learning behavior types that adapt to a specific learning environment. That is, the perceived affordance can be specified through the behavior log data representing the type and level of the behavior.

The eye movement data generated during the learning process is a representation that reflects the learner's cognitive information processing process, and can be an index of perceived affordance. Different types of affordances expressed in various forms can form a discriminatory visual pattern. Gaze fixation and gaze leap show the changing patterns of learners' interests and interests, and the gaze path can visually represent the aspect of attention. have.

In the context of affordance, the movement of the eyeball should not be viewed as an independent movement of the eye, but as a total movement that occurs in association with the head. This is because sight does not passively accept external stimuli, but recognizes information about visual scenes received through the eyes with the head.

In the context of perceived affordance, the learner's cognitive processing process is essential in the act of seeing. The visual stimulus itself from the external environment does not cause the learner to induce meaningful behavior related to learning, but the behavior-inducing characteristic can be expressed through the cognitive processing process inside the learner.

In addition, the learning outcome prediction model derivation apparatus 100 may compare and analyze learning outcome prediction models derived for various groups through such perceived affordance, and provide a personalized, personalized teaching-learning prescription to learners.

2 is a flowchart showing a method of deriving a learning outcome prediction model according to an embodiment of the present invention.

Referring to FIG. 2, in step 210, the apparatus 100 for deriving a learning outcome prediction model may collect behavioral data and physiological psychological data of a learner generated in a learning process for learning content. The learner's behavior data, i.e., behavior log data, can be collected through a video learning player, and at this time, the generated behavior log actions are largely 17 types, and can be re-categorized into 8 types as shown in Table 1 below. have.

<Table 1>

In addition, the physiological and psychological data of the learner, that is, eye movement data generated in the course of learning about the learning content, is "ID" to identify the learner, "LocalTime" to indicate information on the learning time, and information to distinguish between gaze fixation and gaze leap. "GType" representing the eye movement duration, "GDuration" representing the duration of eye movement, and "AOI.A1" to "AOI.B6" representing the set AOI (Area of Interest) information may be included.

In step 220, the learning outcome prediction model derivation device 100 may determine variables for the learner's behavioral data and physiological psychological data collected in the learning process through preprocessing of the collected learner's behavioral data and physiological psychological data. have. In this case, the learning outcome prediction model derivation apparatus 100 may exclude outliers from the collected learner's behavioral data and physiological psychological data.

More specifically, in the case of behavioral data, 1) the data of learners whose all rows are N/A in the original data, that is, learners whose all data are considered meaningless, can be determined as outliers and removed. And, by 2) correcting the time information and action name incorrectly recorded due to a system error, 3) the final data set can be created.

And, in the case of physiological and psychological data (ex. eye movement) 1) Among the data of the original data, data with a ratio of 50% or more deviating from the recommended distance (50-80cm) between the eye tracker and the eyes are viewed as outliers and removed, and measurement software 3) The final data set can be created by removing data whose sum of the validity codes of both eyes (0-8, the lower the value is) exceeding 4 as an outlier.

Thereafter, the learning outcome prediction model derivation apparatus 100 may calculate a variable of the behavior log based on the category of the behavior log action for the generated final data set. Frequency, duration, and regularity can be considered as a representative method of calculating the variables related to the behavior log. Accordingly, the learning outcome prediction model derivation apparatus 100 calculated the frequency and duration for the eight categories of the behavior log actions according to Table 1 below, and the behavior logs corresponding to the final 16 as shown in Table 2 below. Variable types and calculation methods were calculated.

<Table 2>

In the case of the frequency, it can be calculated as the sum of the frequencies of the action log actions that occurred for each category of the action log. The duration may be calculated differently for each category of the behavior log. Specifically, the play action and the pause action may be calculated as the sum of the video playback time and pause time accumulated for each learning video. A comment registration/confirmation action, a playback environment adjustment action, an annotation filtering action, a video viewpoint search/movement action, and a slide movement action may be calculated as a cumulative time required from the time when each action occurs to the time when the next action occurs. In addition, the bookmark registration/confirmation action is a bookmark section actually registered by the learner in the learning video, and may be calculated through the accumulated time based on the video time of the registered bookmark section.

Meanwhile, in order to examine a learner's eye movement in a video-based learning environment, the learning outcome prediction model derivation apparatus 100 may set a specific area of interest (AOI) in a given video-based learning environment. More specifically, the apparatus 100 for deriving a learning outcome prediction model may designate six AOIs for each learning video on the screen of a video learning player, which is a learning medium serving as a video-based learning environment, as shown in FIG. 3.

For example, referring to FIG. 3A, the AOI area 1 is an area in which a learning video is output, and the AOI area 2 is an area in which registered annotations can be checked. AOI area 3 is a library area where registered bookmarks and comments of other learners can be viewed, and AOI area 4 is an area where lecture plans for learning are provided. In addition, referring to FIG. 3B, the AOI area 5 is a pop-up area that is activated when a comment is written, and may not be activated when viewing a learning video.

As such, the variables of eye movement can be calculated based on the set AOI area, and the total duration and average duration for eye fixation and gaze leap, which are representative eye movement calculation methods, can be calculated for each AOI area. At this time, the total duration variable can be calculated as the sum of the duration per episode that occurs for each type of eye movement, and the average duration variable can be calculated by dividing the total duration variable by the frequency of occurrence. The variable types and calculation methods of the eye movements finally calculated as described above may be shown in Table 3 below.

<Table 3>

In addition, the learning outcome prediction model derivation apparatus 100 finally calculates a total of 30 as the variable of the behavior log and the variable of the eye movement, and the variable of the final behavior log and the variable of the eye movement are 7 in the video-based learning. Results categorized based on branch learning activities may be shown in Table 4 below.

<Table 4>

In a later step 230, the learning outcome prediction model derivation apparatus 100 may derive learning outcome prediction models for the entire group and subgroups of the entire group based on the determined learner's behavioral data and variables for physiological and psychological data. have. First, the learning outcome prediction model derivation device 100 performs a descriptive statistical analysis to understand the behavior log and eye movement overall, divides the group according to the task complexity level, and verifies the difference between the behavior log and the eye movement. Each group classified according to the level of task complexity can be classified into detailed clusters through a Gaussian Mixture Model (GMM) cluster analysis.

GMM is a representative algorithm of distribution-based cluster analysis, and it is a cluster analysis method that can obtain analysis results considering the characteristics of various distribution types. GMM assumes that the data subject to cluster analysis are generated by a statistical process and follow different multivariate normal distributions. Since Gaussian is another name for a normal distribution, GMM means that the mean and variance should be estimated with parameters, and the mixed model means that the distributions of several formed models are different.

Since k clusters generated through GMM cluster analysis follow k distributions with different parameters, analysis results can be obtained considering the characteristics of clusters with different distributions. Explaining a cluster by probability distribution means calculating the probability that an individual will belong to a specific cluster and assigning it to the cluster with the largest probability. This allocation method is called soft assignment.

When estimating the parameters of a single distribution from given data, the maximum likelihood estimation method is often used. The method developed to be applied to the mixed model by extending this is the EM (Expectation-Maximization) algorithm, which enables optimal mean and variance estimation. The EM algorithm calculates the probability that the observation belongs to which cluster for each observation value, and assigns each observation value to the cluster with the highest probability.

In GMM cluster analysis, Bayesian information criterion (hereinafter referred to as BIC) is used when setting the number of clusters based on the probability model. Then, after calculating the BIC in consideration of the number of various clusters and the variance structure of each variable, the BIC of each model is compared, and the optimal number of clusters is suggested by referring to the optimal model with the smallest value.

The learning outcome prediction model derivation apparatus 100 is a random forest technique for each of the entire group, the first subgroup divided according to the task complexity level, and the second subgroup divided through cluster analysis of the first subgroup. The model for predicting learning outcomes can be derived by using

Random Forest is an algorithm that applies the CART algorithm of decision tree analysis and the bagging algorithm of the ensemble model. In simple terms, random forest is a model with many (decision making) trees, and it aims to make more accurate predictions by making many decision tree models.

Random Forest constructs several sample datasets by applying the bagging method, and the generated samples and explanatory variables in each node are randomly selected. At this time, the prediction error decreases as the correlation between the decision tree models generated in the random forest decreases, so the best model is derived when the randomness is maximized.

In the random forest, it is possible to generate and analyze OOB (out-of-bag) data without having to attempt model validation by dividing it into training data and test data. This refers to data that is not reselected as a sample when reconstructed (bootstrap) from a random forest, which can be used instead of test data.

Random forest is difficult to interpret because an intuitive model cannot be derived like the graph presented in the decision tree model. As a countermeasure, it is possible to grasp the relative importance (influence) on the response variable of the explanatory variable through a variable of importance index and partial dependence plots of the explanatory variable as a specific number or graph. Particularly, the partial dependence diagram is a diagram proposed to visualize the influence of explanatory variables in an analysis where the number of variables is large and the prediction method is not simple. The relative importance reflected in the partial dependence chart increases the importance index as the influence of the variable increases. Since the importance index follows the standard normal distribution asymptotically, 2 or 3 or more can be the standard.

Finally, the apparatus 100 for deriving a learning outcome prediction model may provide a teaching-learning prescription suitable for a learner through a comparative analysis of changes patterns and important variables of the derived learning outcome prediction models in step 240.

4 is a flowchart illustrating a detailed method of deriving models for predicting learning outcomes for the entire group and detailed groups of the entire group according to an embodiment of the present invention.

Referring to FIG. 4, in step 410, the learning outcome prediction model derivation apparatus 100 performs descriptive statistics analysis on learners' behavior logs, eye movements, and learning outcomes collected in a video-based learning environment. It is possible to check if there are no outliers, and to examine the trends and characteristics of the data through visualization of the behavior log and eye movement. More specifically, descriptive statistics analysis is performed to identify basic forms and patterns of learners' behavior logs, eye movements, and learning outcomes, and can be derived using mean, standard deviation, and maximum/minimum calculation methods.

In step 420, the learning outcome prediction model derivation apparatus 100 may derive a learning outcome prediction model through a behavior log and eye movement using a random forest technique for the entire group. More specifically, the learning outcome prediction model derivation device 100 sets the variables of the final data set generated by preprocessing the collected behavior log and eye movement data as explanatory variables, and predicts the results of the post-test, that is, learning outcomes. By setting it as a variable and applying it to the random forest technique function, a model for predicting learning outcomes can be derived.

In addition, the learning outcome prediction model derivation apparatus 100 may check the results of the learning outcome prediction model and important variables, and check a partial dependence chart for each important variable. At this time, the important variables may be different for each prediction model. For example, in the case of a model for predicting learning outcomes for the entire group, four variables of Play.Freq, BookmarkMTime, MFD.2, and TFD.2 will be identified as important variables. I can.

In step 430, the learning outcome prediction model derivation apparatus 100 may perform a difference test to determine whether there is a difference between the behavior log and the eye movement for each group according to the task complexity level.

First, the learning outcome prediction model derivation apparatus 100 may be divided into individual groups according to the level of task complexity. For example, image 1 (formerly-existing proposition) and image 2 (both term proposition) that the learner learns may be divided into a level of task composition. Specifically, image 1 (formerly-existence proposition) is the prerequisite learning content of image 2 (yanghang proposition) and may have a lower task complexity than image 2.

Task composition, which is the sum of all factors that affect task performance, is affected by the task elements involved, the number of skills, the number of interactions between the elements, and the amount of knowledge required to perform the task element skills. -The existence proposition is the content corresponding to the sub-system of Image 2, and it can be said that the element interaction is low.

On the other hand, the Yanghang proposition covered in Video 2 is a combination of the proposition and the existence proposition, and the number of elements to be understood and the interactions between the elements can be considerably more complex than in Image 1. In addition, the meaning of the proposition is completely different depending on the order of composition of the proposition, and it can be said that the negation of the proposition is also composed of a complex process over several stages, and has a high degree of task complexity.

The learning outcome prediction model derivation device 100 uses a QQ plot and a Shapiro-Wilk test before performing a difference test between groups classified according to the level of task complexity. Normality can be checked, and equal variance can be confirmed through F-test. And, the learning outcome prediction model derivation apparatus 100 may perform a corresponding sample T test for variables that satisfy both normality and equal variance, and for variables that do not satisfy either of normality and equal variance, Wilcoxon code ranking test is performed. Can be done.

In step 440, the learning outcome prediction model derivation apparatus 100 may derive a learning outcome prediction model through a behavior log and eye movement using a random forest technique for the task complex upper and lower groups. At this time, the learning outcome prediction model derivation apparatus 100 may check the results of the learning outcome prediction model and important variables, and check a partial dependence chart for each important variable. Thereafter, the learning outcome prediction model derivation apparatus 100 may compare the learning outcome prediction models of each group and compare important variables and aspects of the learning outcome prediction model according to the level of task complexity.

In step 450, the apparatus 100 for deriving a model for predicting learning outcomes may perform GMM cluster analysis to derive detailed clusters within groups classified according to the level of task complexity. In addition, the learning outcome prediction model derivation apparatus 100 may grasp the characteristics of each group, an action log, and an aspect of eye movement through comparison between the derived detailed clusters.

In step 460, the learning outcome prediction model derivation device 100 is to derive a learning outcome prediction model through behavior logs and eye movements using a random forest technique targeting detailed clusters within groups classified according to the level of task complexity. I can. At this time, the learning outcome prediction model derivation apparatus 100 may check important variables of the results of the learning outcome prediction model and check a partial dependence chart for each important variable. Thereafter, the learning outcome prediction model derivation apparatus 100 may compare the learning outcome prediction models of each cluster and compare important variables and aspects of the learning outcome prediction model according to the level of task complexity.

In addition, the learning outcome prediction model derivation apparatus 100 examines the change pattern of the derived learning outcome prediction model and identifies the relationship between the learner's perceived affordance and the learning outcome in a video-based learning environment, and compares each important variable. The educational applicability of predictive models and appropriate teaching-learning prescriptions can be provided to learners.

The perceived affordance presented in the present invention can be viewed as an expression of the learner's inner process. Therefore, it can be said that the index of perceived affordance visually represents the learner's inner process that appears during the learning process.

Accordingly, the learning outcome prediction model derivation apparatus 100 uses a random forest, a technique that can derive a regression model between variables, with the perceived affordance representation variable as an explanatory variable, and the post-test performance, which is the learning outcome, as an explanatory variable. The relationship between X and Y can be identified.

In addition, important variables may appear differently according to the prediction model derived through the learning outcome prediction model derivation apparatus 100. This is not a simple search for a perceived affordance index that predicts learning outcomes in a linear fashion, but confirms that the variable predicting learning outcomes varies according to the learning content and learner group, and the variable affects the learning outcome. It shows that the influence can be analyzed complexly. Therefore, the fact that important variables vary according to the learner group may imply the necessity of a customized learning outcome prediction model that is subdivided according to the learning content and the learner group.

In addition, the learning outcome prediction model derivation apparatus 100 may provide an appropriate teaching-learning prescription through a customized prediction model subdivided according to the learning content and the learner group as described above.

For example, although the activity categories predicting academic achievement are similar, the specific important variables predicting academic achievement and the patterns of their influence may appear different depending on the level of task composition of the learning content. In the case of the level under the task combination, overall high academic achievement could be predicted according to how much of the function that supports the learning activity was used, whereas in the case of the level of the task combination, not only how the function was used, but also the learner's cognitive process within the learning process. It can be seen that we need to consider how efficiently we use this.

Therefore, the learning outcome prediction model derivation device 100 needs to create a facilitating environment in which learners can actively use the learning medium and easily utilize the provided functions in the learning content with low task complexity, while the task complexity is high. In learning content, it is necessary to foster learners to use the learning environment strategically. In other words, it is required to support a learning experience that effectively encompasses attention to learning content and utilization of the learning environment.

In this case, different perceived affordances may occur depending on the degree of task complexity. In the case of learning content with low task complexity, utilizing the functions provided by the learning environment can promote participation in learning and contribute to improving academic achievement. However, learning content with high task complexity can be perceived by learners. Since affordances can be based on learners' self-directed learning strategies, the learning environment should be designed within a context that supports the learners' self-directed learning processes rather than the functions themselves provided by the environment.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. -A hardware device specially configured to store and execute program instructions such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of the program instructions include not only machine language codes such as those produced by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

The software may include a computer program, code, instructions, or a combination of one or more of these, configuring the processing unit to behave as desired or processed independently or collectively. You can command the device. Software and/or data may be interpreted by a processing device or to provide instructions or data to a processing device, of any type of machine, component, physical device, virtual equipment, computer storage medium or device. , Or may be permanently or temporarily embodyed in a transmitted signal wave. The software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

As described above, although the embodiments have been described by the limited drawings, a person of ordinary skill in the art can apply various technical modifications and variations based on the above. For example, the described techniques are performed in a different order from the described method, and/or components such as a system, structure, device, circuit, etc. described are combined or combined in a form different from the described method, or other components Alternatively, even if substituted or substituted by an equivalent, an appropriate result can be achieved.

Therefore, other implementations, other embodiments and claims and equivalents fall within the scope of the following claims.

100: learning outcome prediction model derivation device

Claims (10)

Collecting behavioral data and physiological psychological data of the learner generated in the learning process for the learning content;
Determining variables for the learner's behavioral data and physiological psychological data collected in the learning process through preprocessing of the collected learner's behavioral data and physiological psychological data; And
Deriving learning outcome prediction models for the entire group and detailed groups of the entire group based on the determined learner's behavioral data and variables for physiological and psychological data
Learning outcome prediction model derivation method comprising a. The method of claim 1,
The deriving step,
Performing descriptive statistical analysis for the entire group based on the determined learner's behavioral data and variables for physiological and psychological data; And
Deriving a learning outcome prediction model for the entire group for which the descriptive statistical analysis was performed using a random forest technique
Learning outcome prediction model derivation method comprising a. The method of claim 2,
The deriving step,
Dividing the entire group into a first subgroup according to a task complexity level by performing a difference test based on the determined learner's behavioral data and variables for physiological and psychological data; And
Deriving a learning outcome prediction model for the first subgroup by using the random forest technique
Learning outcome prediction model derivation method comprising a. The method of claim 3,
The deriving step,
Performing cluster analysis on the divided first sub-groups and classifying the first sub-group into second sub-groups according to characteristics of each group, behavioral data, and physiological psychological data; And
Deriving a learning outcome prediction model for the second subgroup by using the random forest technique
Learning outcome prediction model derivation method comprising a. The method of claim 1,
Providing a teaching-learning prescription to the learner through comparison of the derived learning outcome prediction models by change patterns and important variables
Learning outcome prediction model derivation method comprising a. In the learning outcome prediction model derivation device,
Including a processor,
The processor,
The learner's behavioral data and physiological psychology for the learning process are collected by collecting the learner's behavioral data and physiological psychological data generated in the learning process for the learning content, and preprocessing the collected learner's behavioral data and physiological psychological data. An apparatus for deriving a learning outcome prediction model for determining a variable for data and for deriving a learning outcome prediction model for the entire group and subgroups of the entire group based on the determined variables for the learner's behavioral data and physiological psychological data. The method of claim 6,
The processor,
A learning outcome prediction model that performs descriptive statistical analysis for the entire group based on the determined learner's behavioral data and variables for physiological and psychological data, and derives a learning outcome prediction model for the entire group using a random forest technique. Derivation device. The method of claim 7,
The processor,
By performing a difference test based on the determined learner's behavioral data and variables of physiological and psychological data, the entire group is divided into a first subgroup according to the task complexity level, and the first subgroup using the random forest technique A learning outcome prediction model derivation device that derives a learning outcome prediction model for. The method of claim 8,
The processor,
By performing cluster analysis on the divided first sub-group, the first sub-group is divided into a second sub-group according to the characteristics of each group, behavioral data, and physiological and psychological data, and the random forest A learning outcome prediction model derivation device that derives a learning outcome prediction model for the second subgroup. The method of claim 6,
The processor,
A device for deriving a learning outcome prediction model that provides a teaching-learning prescription to the learner through comparison of the derived learning outcome prediction model change patterns and important variables.

Images