KR100603418B1 - System of providing a learning program and method of providing a learning program using the same - Google Patents

Abstract

The present invention relates to a learning program providing method for providing an optimal learning path for each learner after more accurate evaluation and diagnosis of a learner (subject). The system includes a member identification information database in which the identification information of the learner is stored; According to the attribute data for each item including the item for each level and the corresponding attribute for each item, and the occurrence matrix, the knowledge state matrix, and the abnormal response pattern generated in response to each item by the process of the server according to the attribute data for each item. A test item database configured; A member diagnosis result database for storing test results of each learner of the member identification information database for the questions provided from the test item database; And the databases;

A first step of extracting and outputting a series of problems from the test item database, receiving attribute data and the test result for each learner for the extracted problem, and storing them in the diagnosis result database for each member. After estimating the item parameters according to the item response theory (IRT) of each item through the test results of multiple learners, and then estimating the ability value (θ) of the learner (subject), each learner's ability value is estimated from the estimated result. In a second step of obtaining a heterogeneity coefficient for ζ, a generation matrix indicating whether each attribute is related to each item is generated from the attribute data for each item of the extracted problem in the test item database, and included in the series of problems. Receives the relevant attributes and the number of the generated attributes and generates a knowledge state matrix according to the case / site of each attribute, and accordingly Ability value (θ) and heterogeneity for each abnormal response pattern of the third step are generated by using the third and second step item parameter estimates for generating an abnormal response pattern for the question and storing it in the test item database. Competence value (θ) of each abnormal response pattern obtained in the fourth step from the coordinate value consisting of the subject's ability value (θ) and the heterogeneity coefficient (ζ) obtained in the fourth step and the second step of obtaining the coefficient (ζ). And the fifth step of calculating the Mahallanobis distance of the coordinate value consisting of the heterogeneity coefficient (ζ), the learner (the state of knowledge corresponding to the abnormal response pattern of the recent distance for each learner of the Mahallanobis distance obtained in the fifth step) The sixth step of determining the knowledge state of the test subject and storing it in the diagnosis result database for each member, and the Mahalanenobis distance with each abnormal reaction pattern calculated in the fifth step. Select a group of abnormal patterns within a certain distance, and set the attributes that the learner did not learn to 0, the acquired attributes to 1, and calculate the sum of each attribute to learn the corresponding attributes in the order of the largest sum. And a server performing a process including a seventh step of setting an optimal learning order for each learner and storing it in the prescription result database for each member, thereby providing a problem corresponding to the corresponding optimal learning order. It relates to a learning program providing system and a learning program providing method using the same.

Learning program, ideal response pattern, learner

Description

System of providing a learning program and method of providing a learning program using the same}

1 is a block diagram of a system to which the method of the present invention is applied;

2 is a flowchart illustrating a learning program providing method according to an embodiment of the present invention;

FIG. 3 is a flowchart for explaining the operation from step S16 to step S24 of FIG. 2 in more detail.

※ Explanation of code for main part of drawing

10: user terminal 12: the Internet

14: Server 16: Membership Information Database

18: Test Item Database

20: Diagnostic result database by member

22: Prescription database by member

The present invention relates to a method for providing a learning program, and more particularly, to a method for providing a learning program for each priority to reasonably measure, evaluate, and diagnose learners (subjects) to provide an optimal learning path for each individual.

Recently, with the development of computer technology, the information society has become full-fledged, which has had a great influence on the educational environment. As a result, the computer system is currently being introduced into the educational environment to evaluate and deliver the learning contents. It is expected.

In order for teaching and learning to be effective by computers, subjectivity should be excluded from teaching and learning. In particular, there are many subjective factors that are too subjective to the evaluation of the learners' achievement, which is one of the most important parts of teaching and learning. That is, it is possible to correctly evaluate the learners' ability by making the items to be evaluated for evaluation commercially produced or subject to the subject's subjectivity, and not distinguishing the learners who responded through common photographing, peek, etc. There is no situation.

In addition, most of the learning does not have a device for individual correction, deepening, and supplementation based on a rational theoretical basis, thereby achieving the purpose of learning appropriate for each learner having various skill levels, learning experiences, and attitudes. I can't.

In addition, in the evaluation method, PBT (Paper Based Test) is already implemented as a new method of CBT (Computer Based Test) or CAT (Computer Adaptive Test) method, but measuring the actual ability level of learners is a novice level of measurement evaluation theory. It is evaluated by 100% by the test method, and as a result, diagnosis and prescription depend on arbitrary or error type, which shows the limitation of true learner ability evaluation and diagnosis. Therefore, in order to develop individualized perfect learning and learning ability, which is the ultimate goal of teaching and learning, the cognitive structure considering the cognitive psychology of individual learners, that is, evaluation and diagnosis based on the learner's knowledge state, must be evaluated and diagnosed. You will be able to accurately identify and prescribe weak areas.

The present invention has been proposed in view of the above-described conventional circumstances, and provides a customized learning program for providing an optimal learning path for each learner after evaluating and diagnosing a more accurate intellectual state of the learner (subject). The purpose is to implement the providing system and its method.

In order to achieve the above object, the learning program providing system of the present invention is a learning program providing system using a computer.
The system
Member identification information database that stores the identification information of the learner;
According to the attribute data for each item including the item for each level and the corresponding attribute for each item, and the occurrence matrix, the knowledge state matrix, and the abnormal response pattern generated in response to each item by the process of the server according to the attribute data for each item. A test item database configured;
A member diagnosis result database for storing test results of each learner of the member identification information database for the questions provided from the test item database; And
Including the databases,
A first step of extracting and outputting a series of questions from the test item database, receiving test results for each learner about the extracted problems, and storing the test results in the diagnosis result database for each member, the plurality of learners input from the first step. A second step of estimating the item parameter and each learner's ability value according to the item response theory of each item through the test result of and obtaining the heterogeneity coefficient for each learner's ability value from the estimation result, the extraction of the test item database Generates an occurrence matrix indicating whether each attribute is related to each item from the attribute data for each item related to the problem, and inputs the related attributes and the number of items included in the series of problems. Create a matrix and accordingly generate an abnormal response pattern for the extraction problem A fourth step of obtaining a capability value and a heterogeneity coefficient for each pattern of the abnormal response pattern of the third step by using the third and second item parameter estimates stored in the test item database; And a fifth step of calculating a distance from each pattern of the abnormal response pattern with respect to each learner's coordinates from the coordinate value composed of the capability value and the heterogeneity coefficient obtained from the fourth step, and using the same as a criterion for determining the learner's knowledge state. The server that performs the process
It provides a learning program providing system comprising a.
In addition, the present invention provides a learning program providing method using a computer,
Extracts and outputs a series of questions from a test item database that stores attribute data for each item including questions for each level and corresponding attributes for each item, and receives test results for each learner for the extracted problem. A first step of storing the information in a diagnosis result database for each member according to the stored member identification information database;
A second parameter for estimating the item parameter and each learner's ability value according to the item response theory of each item through the test results of the multiple learners inputted from the first step, and obtaining a heterogeneity coefficient for each learner's ability value from the estimation result step;
From the attribute data for each item related to the extracted problem of the test item database, a generation matrix indicating whether each attribute is related to each item is generated, and the related attributes and the number of the related items included in the series of problems are input and received. A third step of generating a knowledge state matrix according to the case of a site, and generating an abnormal response pattern for the extraction problem and storing it in the test item database;
A fourth step of obtaining a capability value and a heterogeneity coefficient for each pattern of the abnormal response pattern of the third step by using the item parameter estimate of the second step;
A distance from each pattern of the abnormal response pattern with respect to each learner's coordinates from the coordinate values consisting of the capability values and the heterogeneity coefficients obtained from the second and fourth steps, and used as a criterion for determining the learner's knowledge state 5 steps
It provides a learning program providing method comprising a.

Hereinafter, a learning program providing method according to an exemplary embodiment of the present invention will be described with reference to the accompanying drawings.

1 is a configuration diagram of a learning program providing system according to the present invention, which includes the above-described DB, and performs the above-described process in a server. The user terminal 10 can be connected to the Internet 12 as necessary. ; A server 14 for providing a learning program suitable for a learner (subject) using the user terminal 10 to the user terminal 10 connected through the Internet 12; A member identification information database 16 for storing identification information (eg, name, address, ID, password, etc.) of a learner (subject) who has joined as a member; Various types of test items are stored, each test item having a different level of test item database 18; A diagnosis result database 20 for each member in which a diagnosis result determined according to a result of testing and evaluating the learning ability of each member is stored; And a prescription result database 22 for each member that stores a prescription result suitable for each member (that is, a learning program in which a differential learning order is determined for each member) is stored.

Here, the user terminal 10 has a data input means such as a keyboard, a pen mouse or a conventional voice recognition software package, a display means such as a video monitor, a voice output means such as a speaker, and a processing means such as a CPU. It is a computer with a built-in e-mail sending and receiving program and a web-browser. The computer can also be replaced by a laptop, PDA, e-book, or the like.

Then, the server 14 questions and tests a problem of a certain level (CAT: computer adaptive test method) or an arbitrary level in order to test the learner (subject), and the learner (subject) according to the test result We measure the ability of the questionnaire and estimate the item parameter and learner's ability based on the function of the Item Response Theory (three-parameter difficulty, item discrimination, and item estimation). Alternatively, by presenting the question and measuring the ability of the learner (subject) while increasing / decreasing the level of the item according to the response result of the question, the measured value of the corresponding learner (subject) can be estimated. Can be.
Since the estimated value and the response pattern of each learner are based on the response pattern having a difference from the abnormal response pattern according to the knowledge / site of the ideal knowledge, the abnormal response pattern most similar to the response pattern of each learner can be estimated using this. If necessary, the basic score for the learner (subject) is calculated using the true score calculation function of the estimated item characteristic curve for each item, and the partial score is calculated based on the history of the item response. The true score of the learner (subject) is calculated by adding or subtracting the partial score to a basic score, and a predetermined verification (for example, the ability estimation standard error verification and the test information function) is performed on the calculated true score of the learner (subject). Assess the accuracy of the estimation of ability score estimation by It provides a learning program that is optimized for the learning path, but provides a learning program and system optimized for the learner (subject) to learn from the most easily achieved part.

Here, Item Response Theory (IRT) is a theory modeling the relationship between the subject's response and the subject's potential characteristics by probability, and represents the relationship between the subject's response to the item and the potential characteristic by the mathematical function. Is an item response model.

The parameter suggested by the item response theory can be estimated by 1 parameter estimation consisting of item difficulty only, 2 parameter estimation consisting of item difficulty and item discrimination, or 3 parameter estimation consisting of item difficulty, item discrimination, and item estimation. Particularly, in the case of finding the ideal learner's learning level by eliminating the learner's guess or mistake by comparing the difference between the abnormal response pattern and the real learner's response pattern, the three parameter estimation including the item estimation diagram in the estimation This is preferable, and such 1-3 parameter estimation is widely known in theory in the art related to learning theory and statistics.

P (θ) is c + (1-c) * (when the subject with the ability level of θ is P (θ), the probability of answering an item with distinction a, difficulty b and guess c is correct. ^1/1 + e ^{-1.7a (θ-b)} ) can be defined by item response theory.
If the subject's ability value (θ) is plotted on the coordinate plane with the X-axis and the probability P (θ) value of the answer to the question as the Y-axis, the shape of the curve is unique according to the discrimination, difficulty, and guess value of each question. This is called the Item Characteristic Curve.

The item difficulty refers to the point of ability level that the probability of answering the item is 0.5, and the difficulty in the item response model considering the degree of guessing is not the point of the ability level of which the probability of answering the item is 0.5. This is the level of ability that corresponds to the probability at the middle of the highest probability of answering the question.

The item discrimination degree is the degree to which an item is discriminated according to a subject's ability, and is an inclination at the point representing item difficulty on the item characteristic curve.

The item measure indicates the probability that a subject with a level of ability at -∞ will answer the question. The theoretical measure of the item measure index ranges from 0 to 1.0, but is generally less than 1 / (answer). Indicates.

When estimating the item parameters (difficulty, discrimination, and conjecture parameters) of each item included in the test paper, the subject's ability parameters are not known, and only the value of the subjects responding to each item (noon or no answer) can be known. .
First, each item parameter is estimated using the responses of the subjects, and the ability parameter of the subjects is estimated using the estimated item parameters. By repeating the above process several times, it is possible to find the optimal item parameter and the optimal subject ability parameter.
The maximum likelihood estimation method is used in estimating item parameters and ability parameters. The likelihood density function with the unknown term is called the likelihood function. The method of finding the solution of the unknown term that maximizes the likelihood function is called the maximum likelihood estimation method.
The likelihood function in the item parameter and the subject ability parameter estimation becomes an unknown item parameter and the subject ability parameter, and the item parameter and the ability parameter of each subject maximizing the likelihood function are estimated.
Since the solution (item parameter and subject ability parameter) cannot be found in the likelihood function, the solution is found by using Fisher's scoring function, which is an iterative estimation method, and the process is repeated until the estimation value converges. This process is called iteration.
The estimation method is a known method applied in the technical field related to statistics.
Based on the estimated item parameters, item characteristic curves, and ability values, heterogeneity coefficients (ζ) representing the specific degree of each student's item response pattern (response pattern) according to the item response theory for each student's ability value It is possible to express in one index how the type of item response of a student with knowledge status differs from the theoretical item response type of all students with such a level of competence, which is the actual item response type of each subject. The standard heterogeneity coefficient zeta ( ζ ) defined by Tatsuoka (1984) can be calculated by exponentially comparing the difference between the item response type and the ability value and heterogeneity coefficient, respectively, on the x-axis and y. The ability values and heterogeneity coefficients of each learner for each learner can be projected on the xy plane in a two-dimensional plane serving as an axis. The projected space is called a rule book, and the present invention utilizes the space (rule space) to discriminate various types of error solving problems of learners (subjects) and to make mistakes or guesses through statistical classification. The rule space model (RSM) is used by diagnosing and prescribing a true learning state for a pattern of respondents.
The distribution of the projection point by the rule-length model is shown as a bell shape, of which the normal response distribution is called the multinormal distribution and the singular response distribution is called the buck distribution. In the case of a singularity response pattern, it is determined by classifying it as the nearest abnormal pattern, and the distance measurement for finding the abnormality pattern from the singularity reaction pattern may use various conventional distance measurement methods. Mahalanobis Distance (Euclidean distance, which is widely used in geometry) is multiplied by a covariance matrix that takes into account correlations or distributions of data, designed by Indian mathematician Mahalanobis. In this case, the Mahalanobis distance is a distance that takes into consideration the slope of mountain slope and quantitatively expresses the degree of similarity or dissimilarity between the objects. Mahallanobi because you need to measure and evaluate the distance The concept of distance is used, which is already generalized to statistical theories in textbooks rather than publicly known techniques, and it was introduced by Professor Tatsuoka (1987) to discuss the applicability introduced in RSM. do.

delete

The rulebook model statistically classifies the error type of the learner (subject) with the item parameter, the cognitive attribute matrix, the θ parameter, and the ζ parameter.

The rulebook model defines the attributes related to the items, creates the cognitive attribute matrix to grasp the dependencies among the attributes of each item, and applies the Boolean algebra to learner's learning response in the whole set. It simplifies the kind of parameter values of the state of knowledge related to.

When the learner (subject) performs learning according to a predetermined learning program, the server 14 receives the performance result data, measures the achievement, and then displays comments and achievement trends for each performance. Through the transmission to the user terminal 10 of the learner (subject).

Next, a learning program providing operation according to an embodiment of the present invention will be described with reference to the flowcharts of FIGS. 2 and 3.

First, in FIG. 2, when a learner (subject) accesses the server 14 through the Internet 12 using the user terminal 10, the server 14 tests the learner (subject), which learner The initial ability value of the (subject) is set to "0" or randomly assumed, or the learner is tested using the existing ability value data of the existing learner to estimate the ability value of the corresponding learner (subject). At this time, the server 14 extracts an arbitrary problem, or asks the learner (subject) a question that matches the set capability value (step S10).

When the testing for the learner (subject) is finished, the server 14 determines whether the learner (subject) has the ability (θ) by covering the correct answer or the incorrect answer using the correct answer of the problem provided to the learner stored in the server (PBT or CBT), or another method of the capability measurement, suggests an item based on a function value of the three-parameter (item difficulty, item discrimination, item estimation), and raises the level of the item according to the response result of the item. There is a method of measuring (CAT) the ability (θ) of the learner (subject) while turning it down. Here, if a repetitive correct answer to the presented item is input, the item of the upper difficulty is presented, and if a repetitive incorrect answer is input to the presented item, the item of the lower difficulty is presented (step S12).

In the method for measuring the capability, the capability parameter of the subject is estimated by the capability parameter estimation algorithm. According to the item parameter estimation value and the ability parameter estimation value, if the item is more difficult than the subject's ability, it can be judged as the right item by the subject's conjecture, and this is called 'guessing'. On the contrary, if a question having a lower difficulty level than the subject's ability is wrong, it may be judged as a mistake by mistake, and this is called sleeping.

When the evaluation is completed in this way, the server 14 makes a diagnosis based on the evaluation result (step S16). This is described in detail below.

The server 14 prescribes a suitable learning program (learning program having an optimal learning path) to the learner (subject) based on the diagnosis result, and then the user terminal of the learner (subject) through the Internet 12. 10) (Step S18).

Accordingly, the learner (subject) performs the learning by the report order of the learning program displayed on the monitor of his terminal 10 (step S20), and the result of performing the server 14 through the Internet 12. Is sent to.

Accordingly, the server 14 measures the achievement based on the learning performance result of the learner performed by the learner (step S22), and the comments (comments) and achievement trends for each of the learners (subject) through the Internet 12 To the user terminal 10 (step S24). The learner (subject) can accurately grasp the degree of ability by area by looking at the performance-specific commands and achievement trends displayed on the monitor of his terminal 10.

The operation from step S16 to step S24 will be described in more detail with reference to the flowchart of FIG. 3.

When performing the diagnosis based on the evaluation result, the server 14 first prepares to calculate the ability value θ and the heterogeneity coefficient ζ of the learner (subject) (step S100). That is, the ability of the learner (subject) is measured according to the test result, and the item parameter of the item and the learner's ability value are estimated by the function values of the three-parameter (item difficulty, item discrimination, and item estimation). do.

Accordingly, the server 14 generates an occurrence matrix Item-Matrix using the calculated capability value θ and the heterogeneity coefficient ζ (step S110). Here, as an example of the occurrence matrix (that is, defining the problem and the attributes necessary to solve the problem), the following Table 1 can be found.

One 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Attribute 1 0 0 One 0 One 0 One 0 One One 0 One One One One Attribute 2 0 0 0 0 0 0 0 One 0 One One One One One One Attribute 3 0 0 0 One One One One One 0 0 One One One 0 One Attribute 4 0 One 0 0 0 0 0 0 One 0 One 0 One One One Attribute 5 0 0 0 0 0 One One One 0 One 0 One 0 One One Attribute 6 0 0 0 One 0 0 0 One One One One 0 One One One Property7 One 0 One 0 0 One 0 0 One 0 One One One One 0
In the above Table 1, the horizontal axis represents the number of each item provided to the learner, and the vertical axis represents the knowledge attribute required for the solution of the entire item, and indicates whether or not each attribute of the corresponding item corresponds to the " 1 "means the knowledge attribute required for the solution of the question," 0 "means the knowledge attribute that is not necessary for the solution of the question. Accordingly, the server 14 uses the attributes to determine which of the attributes 1 to 7 is unknown depending on whether or not the student solves the problem.

delete

The server 14 generates a matrix of knowledge states of the learner (subject) based on the generation matrix, and generates an ideal response pattern according to the knowledge states as shown in Table 2 below (step S120). That is, if the knowledge related to the whole problem is divided by attribute, all 7 possible knowledge states are 2 ⁷ = 128 since each knowledge attribute has land / site, and if the learner's answer is mistake or guess If not included, the pattern of the answer for each knowledge state will represent only the adverse reaction pattern of Table 2 below.

      State of knowledge         Adverse reaction pattern     One    0,0,0,0,0,0,0,    0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,     2    1,0,0,0,0,0,0,    0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,     3    0,1,0,0,0,0,0,    0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,     . . .         . . .              . . .    127    0,1,1,1,1,1,1,    1,1,0,1,0,1,0,1,0,0,1,0,0,0,0,    128    1,1,1,1,1,1,1,    1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,

2

3

4

5

위 5문제는 속성 A1, A2, A3과 연관이 있다. 즉, 1번 문제를 풀기 위해서는 A2, A3, 2번 문제는 A1, A2, A3, 3번 문제는 A1, A2, 4번 문제는 A2, 5번 문제는 A3을 알아야 한다. 이런 관계를 아래 표 4의 발생 행렬(Att-Matrix)로 표현할 수 있다.
문제 속성 문제 1 문제 2 문제 3 문제 4 문제 5 속성 1 0 1 1 0 0 속성 2 1 1 1 1 0 속성 3 1 1 0 0 1
또한 5개 문제에 대하여 한 학습자가 실제적으로 반응할 수 있는 반응패턴의 총 경우의 수는 5문제 모두를 틀리는 (0,0,0,0,0)로부터 5문제 모두를 맞히는 (1,1,1,1,1)까지 총 32(=2⁵)가지이나, 실제로 문항과 관련된 속성의 획득여부에 따라서는 학습자의 지식상태 행렬은 속성의 이해여부에 따라 3가지 속성을 모두 모르는 (0,0,0)에서부터 3가지 속성을 모두 아는 (1,1,1)까지 지식상태의 수는 8(=2³)가지이다.
번호 지식상태 지식상태 행렬 이상반응패턴 1 A1과 A2와 A3를 모르는 사람 (0, 0, 0) (0,0,0,0,0) 2 A1만 알고 있는 사람 (1, 0, 0) (0,0,0,0,0) 3 A2만 알고 있는 사람 (0, 1, 0) (0,0,0,1,0) 4 A3만 알고 있는 사람 (0, 0, 1) (0,0,0,0,1) 5 A1과 A2만 알고 있는 사람 (1, 1, 0) (0,0,0,1,0) 6 A1과 A3만 알고 있는 사람 (1, 0, 1) (0,0,0,0,1) 7 A2와 A3만 알고 있는 사람 (0, 1, 1) (1,0,0,1,1) 8 A1과 A2와 A3를 알고 있는 사람 (1, 1, 1) (1,1,1,1,1)
또한 상기 8가지의 지식 상태에 대하여 각 문제가 가지는 속성에 따라 각 문제를 풀 수 있는지 여부를 상기 1~5번 문제에 대하여 나타내면 상기 표 5의 이상반응패턴이 구해진다.
따라서 원래 한 피험자가 5문제에 대하여 반응하는 경우는 32(=2⁵)가지이지만, 지식상태로 분류하면 지식상태 1과 2, 3과 5, 및 4와 6은 동일한 반응을 가지는 경우이므로 정상적인 반응을 한 피험자는 32가지 경우의 수 중, 단지 5가지의 반응패턴을 보여야 한다(상기 표 5의 번호1, 3, 4, 7 및 8에 해당하는 패턴). 따라서 상기 5가지 반응패턴을 이상반응패턴(Ideal Response Pattern)이라 하고, 나머지 27가지 반응의 경우는 피험자의 지식상태에 대한 정확한 반응이 아니다. 이러한 이상반응패턴에 일치하지 않는 반응패턴들은 "특이한 반응(Unusual of Fuzzy)"으로 규칙의 일관되지 않는 적용에 따른 찍기 등의 "추측(guessing)" 또는 "실수(slippping)"에 따른 것이다.
그리고 나서, 그 서버(14)는 학습자(피험자)의 반응과 이상적 반응패턴과의 거리를 마할래노비스 계산방법을 통하여 계산하여 가장 최근 거리를 갖는 적합한 지식상태를 찾아낸다(단계 S140). 즉, 다음의 표 6에 학습자(피험자)의 반응패턴(표 6 上)의 마할래노비스 거리가 최근 거리 이상반응패턴을 찾게 되면(표 6의 下 지식상태 14), 그 이상반응패턴에 해당하는 지식상태를 학습자(피험자)의 지식상태로 진단한다(단계 S140).Subsequently, the server 14 may calculate the capability value θ and the heterogeneity coefficient ζ for each of the ideal response patterns by using the result value of the item parameter estimate, and coordinates for each of the abnormal response patterns. The value is displayed in the rule book having the capability value and the heterogeneity coefficient as the x-axis and y-axis, respectively (step S130). In other words, the calculated heterogeneity coefficient ζ measuring the specific value of the capability value θ and the response pattern is shown in a rectangular coordinate consisting of two axes, and the rule and response type are represented as projection points.
For example, in the inspection consisting of five problems as shown in Table 3, the attributes A1, A2, and A3 are defined as follows, respectively. That is, attribute 1 (A1) relates to knowledge that can represent two fractions with different denominators as a common denominator, and attribute 2 (A2) relates to knowledge that can calculate the sum of two fractions with the same denominator. For example, attribute 3 (A3) relates to knowledge that can represent fractions in the form of a single fraction.
number Problem One

2

3

4

5

The above 5 problems are related to attributes A1, A2 and A3. That is, in order to solve problem 1, A2, A3, problem 2 should be A1, A2, A3, problem 3 is A1, A2, problem 4 is A2, problem 5 is A3. This relationship can be expressed by the generation matrix (Att-Matrix) of Table 4 below.
Problem attribute Issue 1 Problem 2 Issue 3 Issue 4 Issue 5 Property 1 0 One One 0 0 Property 2 One One One One 0 Property 3 One One 0 0 One
In addition, the total number of response patterns that a learner can actually respond to five questions is (1,1, which matches all five questions from (0,0,0,0,0) which is incorrect in all five questions. A total of 32 (= 2 ⁵ ) up to 1,1,1), but depending on whether or not the attribute related to the item is actually obtained, the learner's knowledge state matrix does not know all three attributes depending on whether the attribute is understood (0,0). From (0) to (1,1,1) where all three properties are known, the number of knowledge states is 8 (= 2 ³ ).
number State of knowledge State of knowledge matrix Adverse reaction pattern One Persons who do not know A1, A2 and A3 (0, 0, 0) (0,0,0,0,0) 2 A person who only knows A1 (1, 0, 0) (0,0,0,0,0) 3 A person who only knows A2 (0, 1, 0) (0,0,0,1,0) 4 A person who only knows A3 (0, 0, 1) (0,0,0,0,1) 5 A person who only knows A1 and A2 (1, 1, 0) (0,0,0,1,0) 6 A person who only knows A1 and A3 (1, 0, 1) (0,0,0,0,1) 7 A person who only knows A2 and A3 (0, 1, 1) (1,0,0,1,1) 8 Person who knows A1 and A2 and A3 (1, 1, 1) (1,1,1,1,1)
In addition, the problem response patterns of Table 5 can be obtained by indicating whether the problems can be solved according to the property of each problem for the eight knowledge states with respect to the first to fifth problems.
Therefore, one subject responds to 5 problems in 32 cases (= 2 ⁵ ), but when classified as knowledge state, knowledge states 1 and 2, 3 and 5, and 4 and 6 have the same reaction, so they are normal reactions. Subjects should show only five response patterns out of 32 cases (patterns corresponding to numbers 1, 3, 4, 7 and 8 in Table 5 above). Therefore, the five response patterns are referred to as an ideal response pattern, and the remaining 27 reactions are not accurate responses to the knowledge state of the subject. Response patterns that do not correspond to these abnormal reaction patterns are "Unusual of Fuzzy" and are according to "guessing" or "slippping", such as photographing due to inconsistent application of the rule.
Then, the server 14 calculates the distance between the learner's (subject) response and the ideal response pattern through the Mahallanobis calculation method to find a suitable state of knowledge having the most recent distance (step S140). That is, in the following Table 6, when the Mahallanobis distance of the learner's response pattern (above Table 6) finds the recent distance abnormal response pattern (below knowledge state 14 in Table 6), the corresponding abnormal response pattern The knowledge state is diagnosed as the knowledge state of the learner (subject) (step S140).

delete

student q1 q2 q3 q4 q5 q6 q7 q8 q9 q10 q11 q12 q13 q14 q15      θ     ζ A One 0 0 0 One 0 0 0 0 0 0 0 0 0 0 -1.42197732 0.06086458

  State of knowledge       Adverse reaction pattern    theta (θ)     zeta (ζ)  14  1,0,1,1,0,0,0,  0,1,0,0,1,0,0,0,0,0,0,0,0,0,0   -1.40679432   0.13818616  16  1,1,1,1,0,0,0,  0,1,0,0,1,0,0,0,0,0,0,0,0,0,0   -1.40679432   0.13818616

The server 14 determines the learning order of the attribute to be learned in the knowledge state of the diagnosed learner (step S150). That is, the server 14 clusters and selects abnormal response patterns in which the Mahallanobis distance is close to a predetermined distance from the learner's ability value and heterogeneity coefficient coordinates for each learner in the rule book (RSM). As shown in Table 7 below, the knowledge states corresponding to the abnormal reaction patterns can be represented, and the learning order is determined from the most easily achieved part by the result of calculating the sum (SUM) for each attribute for each knowledge state (SUM). Step S150).

   A1    A2    A3    A4    A5    A6    A7    One    One    One    0    One    One    One    0    2    One    0    One    One    0    0    One    3    0    0    One    0    One    One    One    4    One    0    0    0    0    One    One    5    One    0    0    0    One    One    One    6    One    One    One    0    0    0    One   SUM    5    2    3    2    3    4    5

For example, when the estimated learner's knowledge state is "1,0,1,1,0,0,0" in Table 6, the six knowledge states that are close to each other within a certain distance are 1 (1) in Table 7. , 1,0,1,1,1,0), 2 times (1,0,1,1,0,0,1) to 6 times (1,1,1,0,0,0,1) S14, A3, and A4 in the knowledge state of the learner, since the learner has already acquired the knowledge, the knowledge attribute to be learned by the learner (property corresponding to slip) A2, A5, A6, and A7. In step S150, the order of attributes to be learned is first determined. That is, in Table 7, feedback is provided to the learners (subjects) in the order A7 → A6 → A5 → A2, in which the sum of the attributes is larger in the clustered pattern among the attributes (step S160).

Accordingly, the learner (subject) learns the questions provided in order, and when the learner (subject) performs the learning according to a predetermined learning program, the result data is provided through the Internet 12. Achievement is measured (step S170). The measurement of the achievement means that a known learner confirms a result of performing learning according to the predetermined learning program, which is transmitted through a user terminal connected to the server or outputted by the server to the learner. The method may be applied, and an example thereof may be a method through a web browser of a user terminal.

Thereafter, the server 14 transmits the performance-specific comments and achievement trends of the learning performance result to the user terminal 10 of the corresponding learner (subject) through the Internet 12. The achievement-specific commands and achievement trends refer to various methods used in a known electronic learning management system, which may be applied through a user terminal connected to the server or a method of sending a result output from the server to a learner. For example, on the monitor of the user terminal 10 of the learner (subject), the achievement trend indicating a change in the results of the achievement of the grades and weakness attributes according to the predetermined learning program and the resulting results Is displayed (step S180).
In another aspect, the present invention provides a learning program providing method proceeded by the system described above. The learning program providing method means a process including the above-described steps as a learning program providing method using a computer.

As described in detail above, according to the present invention, the learning ability of the learner (subject) can be accurately and objectively evaluated, and customized education can be made, and in particular, learner-centered custom education is possible.

In addition, it is possible to tailor the education to meet the learner's ability level, it is possible to present a step-by-step learning program and to develop textbooks.

In addition, the learner can accurately grasp the degree of ability of each area, it is possible to select a career that suits the ability.

In addition, online learning is assessed, diagnosed, and prescribed, so that equal learning and evaluation opportunities beyond time and local limits are granted.

In addition, intensive learning in vulnerable areas not only improves performance, but also doubles the adaptability to new learning environments.

On the other hand, the present invention is not limited only to the above-described embodiment, but can be modified and modified within the scope not departing from the gist of the present invention, the technical idea to which such modifications and variations are also applied to the claims Must see

Claims (12)

In the learning program providing system using a computer, The system Member identification information database that stores the identification information of the learner; According to the attribute data for each item including the item for each level and the corresponding attribute for each item, and the occurrence matrix, the knowledge state matrix, and the abnormal response pattern generated in response to each item by the process of the server according to the attribute data for each item. A test item database configured; A member diagnosis result database for storing test results of each learner of the member identification information database for the questions provided from the test item database; And Including the databases, A first step of extracting and outputting a series of problems from the test item database, receiving attribute data and the test result for each learner for the extracted problem, and storing them in the diagnosis result database for each member. Through the test results of multiple learners, the item parameters according to the item response theory of each item were obtained by MML and EL estimation, and the learner's ability values were obtained by ML, MAP, or EAP estimation. A second step of obtaining a heterogeneity coefficient for a value, generating an occurrence matrix indicating whether each attribute is related to each item from the attribute data for each item of the extracted problem in the test item database, Generates the knowledge status matrix according to the land and site of each property by inputting the property and number And generating an abnormal response pattern corresponding to the extraction problem and storing it in the test item database using the estimated parameter values of the third and second steps. (θ) and the heterogeneous coefficient ζ of the fourth step, from the coordinate value consisting of the subject's ability value (θ) and heterogeneity coefficient (ζ) obtained in the second step of each of the adverse reaction pattern obtained in the fourth step The fifth step of calculating the Mahallanobis distance of the coordinate value consisting of the ability value (θ) and the heterogeneity coefficient (ζ), and the Mahallanobis distance obtained in the fifth step corresponds to the abnormal response pattern of the recent distance for each learner. The sixth step of determining the knowledge state as the knowledge state of the learner (subject) and storing it in the diagnosis result database for each member, and with each discrete response pattern calculated in the fifth step From the Mahallanobis distance, select the group of abnormal patterns within a certain distance, and set the attribute that the learner did not learn in the abnormal pattern group to 0, the acquired attribute to 1, and sum each attribute. A server that performs a process comprising a seventh step of setting the optimal learning order for each learner to learn the corresponding attribute and storing it in the prescription result database for each member, thereby providing a problem corresponding to the corresponding optimal learning order. Learning program providing system comprising a. The method of claim 1, The learning program providing system further includes a learner's user terminal connected to the server by a network, wherein the output of the problem of the first step and the input of a test result or the provision of the corresponding problem of the seventh step are performed by the user terminal. Learning program providing system, characterized in that made through. The method of claim 1, The item parameter of the second step is a three-parameter consisting of item difficulty, item discrimination, and item estimation, The difficulty level of the item is a point of ability level with a probability that the answer of the item is 0.5. The item discrimination degree is the degree to which the item is discriminated by the learner according to the ability, the learning program providing system, characterized in that the slope in the point representing the item difficulty on the item characteristic curve. The method of claim 1, The distance calculation of the fifth step is a learning program providing system, characterized in that to apply Mahallanobis distance measuring method. The method of claim 1, The diagnosis result database for each member further includes a true test result of the learner adjusted by performing the process of the server, Further comprising a member-specific prescription results database for storing the optimal learning order for each learner generated by performing the process of the server from each member's true test results of the member-specific diagnostic result database, In the above process, The sixth step of selecting the abnormal response pattern of the recent distance for each learner from the distance obtained in the fifth step to determine this as a true test result of the learner and to store it in the diagnosis result database for each member, and the calculated in the fifth step The group of abnormal patterns within a certain distance is selected from the distance from each pattern, and the attributes not acquired by the learner in these abnormal pattern groups are set to 0, the acquired attributes are set to 1, and the sum of each attribute is calculated, and the sum is high. The seventh step of setting the optimal learning order for each learner to learn the corresponding attribute to store the information in the prescription results database for each member, thereby providing a problem corresponding to the optimal learning order Learning program providing system further comprises a. The method of claim 5, And repeating the first to seventh steps with respect to the test result of the problem provided in the seventh step in the process, and providing each learner with a trend of improving the learning ability according to the repetition. Learning program providing system. In the learning program providing method using a computer, Learner extracts and outputs a series of questions from a test item database that stores item-specific attribute data including questions for each level and corresponding attributes for each item, and inputs the attribute data and the test result for each learner. A first step of storing the personal information of the member in a diagnosis result database for each member according to the member information database where the personal information is stored; Through the test results of the multiple learners inputted from the first step, the item parameters according to the item response theory of each item are obtained by MML and EL estimation, and the learner's ability values are obtained by ML, MAP, or EAP estimation. Obtaining a heterogeneity coefficient for each learner's ability value from the result; From the attribute data for each item related to the extracted problem of the test item database, a generation matrix indicating whether each attribute is related to each item is generated, and the related attributes and the number of the related items included in the series of problems are input and received. A third step of generating a knowledge state matrix according to the case of a site, and generating an abnormal response pattern for the extraction problem and storing it in the test item database; A fourth step of obtaining the ability value θ and the heterogeneity coefficient ζ for each of the abnormal reaction patterns of the third step by using the item parameter estimation value of the second step, and the subject's ability value obtained in the second step ( a fifth step of calculating a Mahalanobis distance of the coordinate value composed of the heterogeneity coefficient ζ and the capability value θ of each of the abnormal reaction patterns obtained in the fourth step from the coordinate value composed of θ) and the heterogeneous coefficient ζ. In a sixth step of determining the knowledge state corresponding to the abnormal response pattern of the recent distance for each learner from the Mahallanobis distance obtained in the fifth step as the knowledge state of the learner (subject) and storing it in the diagnosis result database for each member , And select a group of abnormal patterns within a certain distance from the Mahallanobis distances with the discrete reaction patterns calculated in the fifth step, and learners did not learn in these abnormal pattern groups. Is 0, the acquired attribute is 1, the sum of each attribute is obtained, and the optimal learning order for each learner is set to learn the corresponding attribute in the order of the largest sum, and the result is stored in the prescription result database for each member. A learning program providing method comprising the seventh step of providing a problem corresponding to the corresponding optimal learning order. The method of claim 7, wherein In the learning program providing method, the learning program providing method is executed in a server, and the user terminal connected to the server by outputting a problem and outputting a test result of the first step or providing a corresponding problem in the seventh step is networked. Method for providing a learning program, characterized in that made through. The method of claim 7, wherein The item parameter of the second step is a three-parameter model consisting of item difficulty, item discrimination, and item estimation. The method of claim 7, wherein The distance calculation of the fifth step is a learning program providing method, characterized in that for applying the Mahallanobis distance measuring method. The method of claim 7, wherein A sixth step of selecting an abnormal response pattern of the most recent distance for each learner among the distances obtained in the fifth step, determining this as a true test result of the learner, and storing it in the diagnosis result database for each member; And The group of abnormal patterns within a certain distance is selected from the distances with the patterns calculated in the fifth step, and the attributes not acquired by the learner in these abnormal pattern groups are set to 0, and the acquired attributes are set to 1, The seventh step of setting the optimal learning order for each learner to obtain the sum and learning the corresponding attributes in the order of the largest sum and storing it in the prescription result database for each member, thereby providing a problem corresponding to the corresponding optimal learning order. Learning program providing method characterized in that it further comprises. The method of claim 11, Repeating the process of the first to seventh step with respect to the test result of the problem provided in the seventh step and providing a learner improvement trend according to the repetition to each learner How to provide a learning program.

Images