Online learning user matching method based on self-study completion index

Abstract

The present application relates to the field of online accompanying self-study, in order to solve how to match the self-study completion condition with the current user similar system user, so that the user between the matching self-study promotes each other, the present application proposes an online accompanying learning user matching method based on self-study completion index, including the following steps: self-study completion index grade point initialization of user; correction is carried out to the matching grade point based on self-study completion index grade point; in the interval of the matching target self-study completion index grade point after correction, user matching is carried out; repeat user matching and self-study completion index grade point correction; after repeating matching for several times, the self-study completion index grade point of the user will tend to be stable after matching correction, and the self-study completion index of the user is obtained.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of online accompanying self-study, in particular to an online accompanying self-study user matching method based on a learning completion index. BACKGROUND

[0002] With the development of the Internet and the normalization of epidemic prevention, online self-study is becoming more and more popular, and online accompanying self-study is conducive to mutual supervision, mutual accompanying, and efficient learning.

[0003] The mode of online accompanying self-study is as follows: a user sets a self-study goal in the system, including a self-study duration; the system records the signal and immediately matches the user with other users with a duration as close as possible to the user's self-study duration, and the two users self-study together before the self-study duration set by the user expires; when the self-study duration expires, the user sends a self-study end signal to the system, and the system ends the self-study of the two users.

[0004] The problem of online accompanying self-study is that the matched users may have high self-discipline and be able to complete the self-study according to the set goal, or may be matched with users who cannot complete the self-study goal, and the completion of the self-study is also related to the user's own learning habits and ability to some extent. When a user is matched with a user who has a large difference in self-study completion, the one who focuses on completing the self-study goal may be affected by the other, thereby affecting the effect of online accompanying self-study of the two parties. Matching users with similar self-study completion to the current user together can to some extent reduce the gap between the two parties in terms of self-study habits, self-discipline, etc., and thus better utilize the psychological mechanism of wanting to quit self-study earlier than the other party, promote the matched users to better complete their set self-study goals, and thus help users gradually improve their self-study completion. SUMMARY

[0005] In order to solve the problem of how to match users with similar self-study completion to the current user, thereby promoting and supervising the users matched for self-study, the present application proposes

[0006] An online accompanying self-study user matching method based on a self-study completion index, comprising the following steps:

[0007] S1. Initialize the self-study completion index grade of each user in each learning period:

[0008] For a user who newly enters the online accompanying self-study system or a user whose total number of matching sessions in the last learning period is less than ten, the self-study completion index grade is initialized as:

[0009]

[0010] Wherein, μ is the overall average self-study completion index rating of the existing users of the online self-study system, and σ is the standard deviation; for the old users of the online self-study system whose total matching fields in the last learning period are not less than ten, the self-study completion index rating is initialized by the following formula:

[0011] R p = R c + D(P)

[0012] Wherein: R p is the predicted rating of the current learning period; R c is the average rating of all users matched in the last period; P is the historical self-study completion rate of the current user in the last period.

[0013] S2. The matching rating based on the self-study completion index rating is corrected,

[0014] When matching the user, a rating for matching the appropriate object for the current user is obtained based on the self-study completion index rating, i.e. the matching target self-study completion index rating.

[0015] The self-study completion index rating formula is as follows:

[0016] R n = R o + (W s -W e )

[0017] Wherein: R n is the new rating after matching; R o is the original rating before matching; W s is the actual self-study completion, which is 1 when the self-study completion of the own side is completed and the completion of the other side is not completed, 0.5 when both sides are completed, and 0 in other cases; W e is the expected self-study completion, which is equal to the expected self-study completion rate.

[0018] If the user continuously completes the self-study target, the matching target self-study completion index rating is corrected before each matching by the following formula,

[0019] R m = R′ + R d

[0020]

[0021] Wherein, R m is the matching target self-study completion index rating; R′ is the average self-study completion index rating of all users in the continuous self-study field; R dTo match the modified score, σ is the standard deviation, n is the number of consecutive completed self-study in the interval of the grade partition, and c is a constant;

[0022] In other cases, the matching target self-study completion index grade partition is equal to the current user's self-study completion index grade partition;

[0023] R m = R

[0024] Where R m is the matching target self-study completion index grade partition, and R is the current user's self-study completion index grade partition;

[0025] S3. In the interval of the grade partition where the modified matching target self-study completion index grade partition is located, matching is performed:

[0026] In the interval of the grade partition where the user's self-study completion index grade partition is located, the user with the closest self-study completion index score is preferentially found for matching with the same self-study duration set this time;

[0027] S4. After the matching of both parties ends, the new self-study completion index grade partition of both parties is calculated through the current grade partition formula;

[0028] R n = R o +C×K×(W s -W e )

[0029]

[0030] Where: R n is the new grade partition after matching; R o is the original grade partition before matching; W s is the actual self-study completion, which is 1 when the self-study completion of one party is not completed by the other party, 0.5 when both parties complete it, and 0 otherwise; W e is the expected self-study completion, which is equal to the expected self-study completion rate;

[0031] K is an uncertainty function; C is a compensation function, D is the difference between the grade partition values of the other party and oneself, and σ is the standard deviation;

[0032] S5. Repeat steps S2-S4 several times:

[0033] Determine whether the user's self-study completion index value is stable by the maximum-minimum difference method.

[0034] As an option, in step S3, the specific method for preferentially finding the user with the closest self-study completion index score for matching in the interval of the grade partition where the user's self-study completion index grade partition is located is as follows:

[0035] According to the optional planned learning time length, the users in the match are divided into several match sub-queues with length ≤N, wherein N is the maximum length of each match sub-queue;

[0036] In each sub-queue, according to the user request order, the first come first served (FCFS) algorithm is used to match the users one by one from the user who sends the match request first, the quick sort is used to find the user with the most similar self-study completion index value, and the user is removed from the queue until the match request of all users in the queue is satisfied.

[0037] When a sub-queue is full, the new match request is placed in the waiting queue corresponding to the sub-queue; when the sub-queue is not full, the match request is inserted into the corresponding match sub-queue in order; if the user match in a sub-queue appears timeout, it means that there is only one user in the queue, at this time the user is transferred to the non-empty sub-queue with the closest target self-study time length, and the matching is repeated in step S3 after the transfer.

[0038] Repeat step S3 until the match request of all users in the queue is satisfied, at this time there may still be a request of the last user that cannot be satisfied, at this time the system prompts that the matching fails, and the matching ends.

[0039] As a preferred, in the step S3, the range of the grade partition is The following, Wherein, μ is the overall average self-study completion index grade of the existing users of the online accompanying self-study system, and σ is the standard deviation.

[0040] As a preferred, in the S5, the specific method for judging whether the user self-study completion index value is stable by the maximum difference method is as follows:

[0041] When the total number of fields in the current learning period of the user is greater than ten, the system will obtain the maximum value Max, the minimum value Min and the average value Avg of the user self-study completion index value according to all the match records in a week and the latest ten match records,

[0042] If (Max-Min) / Avg<5%, it is considered that the value tends to be stable, the cycle is stopped, and the real self-study completion index of the user in this learning period is obtained, which is used for more accurate user matching.

[0043] The present application takes the normal distribution of the learning individual ability as the breakthrough point, and proposes a self-study completion index rating scale suitable for learning individuals, which provides a basis for interval division of learning matching. And, taking the logistic function as the prototype, the normal cumulative distribution function is optimized through reasonable scaling, so as to realize the estimation of the self-study target completion of each matching of the learning individual. Through the matching method proposed by the present application, after several times of matching, the user's self-study completion index can be numerized by means of the current rating formula, and the purpose of matching individuals with similar self-study completion indexes is achieved. Further, in order to shorten the matching process as much as possible and optimize the learning experience, the present application proposes an uncertainty function and a matching target rating correction function which can be searched and planned in a grid. In addition, in the process of long-term persistence of learning matching, the mutual supervision and mutual learning between users can also improve the learning consciousness, concentration and learning habits of the users, so as to improve the self-study completion rate.

[0044] The effectiveness of the above-mentioned online accompanying learning user matching method based on the self-study completion index is demonstrated as follows: the self-study completion of different individuals can be objectively obtained from system data, and the analysis of system data shows that the self-study completion of different individuals basically conforms to the normal distribution, and the average self-study completion of individuals tends to be stable in the same time period, and the proportion of large changes is not high, so the difference in self-study completion between individuals can be expected according to the historical self-study completion of each user. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or prior art description. The specific embodiments of the present application will be described in detail below with reference to the drawings in an exemplary but not limiting manner. The same reference signs in the drawings denote the same or similar components or parts. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings:

[0046] Figure 1 Flowchart for self-study completion index numerization. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0048] In normal distribution, the probability of x falling outside of (μ-3σ, μ+3σ) is less than 1 / 1000, so we can divide the individual ability values into eight intervals, μ-3σ below, (μ-3σ, μ-2σ), (μ-2σ, μ-σ), (μ-σ, μ), (μ, μ+σ), (μ+σ, μ+2σ), (μ+2σ, μ+3σ), μ+3σ above. Obviously, we can consider that the learning individuals in the same interval have similar learning ability, so we can match them together.

[0049] Further, since μ is the average individual ability value, in order to avoid the problem of frequent cross-interval matching, we can move the above intervals to the left That is The following, The above.

[0050] Take μ=1700, σ=200 as an example: 1200 below, (1200, 1400), (1400, 1600), (1600, 1800), (1800, 2000), (2000, 2200), (2200, 2400), 2400 above.

[0051] Table 1-1 Grade interval and ability comparison table

[0052]

[0053]

[0054] Where P1(D) is a modified function of the standard normal cumulative distribution function, D is the difference between the two parties, and σ is the standard deviation.

[0055] The mathematical meaning of P1(D) is to take the difference between the current user and the other party's self-study completion index grade points as the upper limit variable of the integral, so as to obtain the probability of observing the other party within a certain interval between the baseline and the standard normal distribution curve. And the probability of mutual observation of both parties is 1. In practical terms, the difference between P1(D) of both parties means the difference between the self-study completion indexes of both parties; the party with a higher P1(D) obviously has a relatively higher self-study completion index and a better self-study completion situation. Further, its property of adding up to 1 means that the size of P1(D) of the current user is the expected situation of the current user completing self-study, i.e. the self-study completion rate, under the mutual influence of both parties. Therefore, the P1(D) function is mainly used to estimate the individual self-study completion rate according to the difference between the matching parties' self-study completion index grade points. According to the assumed premise, the self-study completion rate describes the situation of the learning individual completing the self-study goal. Since the function calculation is not convenient and the logistic function and the normal cumulative distribution function have very similar images, and the two variables have no different meaning in practical application, the standard logistic function is used to replace the representation of the standard normal cumulative distribution function.

[0056] The logistic function is as follows

[0057]

[0058] Where: x0 is the value of x taking the midpoint of sigmoid; L is the maximum value of the curve; k is the logistic growth rate or the steepness of the curve. The standard logistic function is as follows

[0059]

[0060] We can use to approximate e, and then scale it according to σ for easy calculation. For example:

[0061]

[0062] Where: D represents the difference between the two parties.

[0063] When applied to practical problems, when σ = 200, if the grade point score of A is x a and the grade point score of B is x b , i.e.

[0064]

[0065]

[0066]

[0067] Where: Eb Self-study completion rate expectation for B; E a Self-study completion rate expectation for A.

[0068] Further, it is not difficult to get

[0069] E a + E b = 1 (3-8)

[0070] And, taking B as an example, the scaling means that if B's score is 400n more than A's, then B's self-study completion rate expectation will be 10 n times that of A. Accordingly, we can get the table of P(D) and its inverse function D(P) as follows.

[0071] Table 1-2: Find the self-study completion index grade difference from the score rate

[0072]

[0073] Table 1-3: Find the self-study completion rate from the self-study completion index grade difference

[0074]

[0075]

[0076] The calculation formula of the self-study completion index expected grade is as follows:

[0077] R p = R c + D(P) (4-1)

[0078] Where: R p is the expected grade of the current learning period; R c is the average grade of all users matched in the last period; P is the historical self-study completion rate of the current user in the last period.

[0079] This formula is used to initialize the grade of users who have been matched for more than ten times in the last learning period, i.e., the grade initialization of old users, in each learning period. If the number of total matching fields of a user in the last learning period is less than ten or the user is a new user, the grade of the user is initialized to μ-2.5σ. Taking μ=1700 and σ=200 as an example, the value is 1200.

[0080] When both self-study ends, the self-study completion index grade of both parties needs to be corrected according to their self-study completion situation using the current self-study completion index grade formula. The current self-study completion index grade formula is as follows:

[0081] R n = R o + (Ws -W e ) (5-1)

[0082] wherein: R n is the new grade point after matching; R o is the original grade point before matching; W s is the actual self-study completion condition (1 if the self-study is completed by the subject and not completed by the opponent, 0.5 if both the subject and the opponent complete the self-study, and 0 otherwise); W e is the expected self-study completion condition, which is equal to the expected self-study completion rate.

[0083] Such a calculation method can gradually converge to the true self-study completion index of the learning individual, although it is matched for a sufficient number of times. However, the convergence speed is obviously very slow and does not meet the application requirements of actual learning matching. Therefore, we can introduce an uncertainty function K to improve the convergence speed. Obviously, the lower the self-study completion index grade point of an individual, the greater the progress space; the higher the self-study completion index score of an individual, the more difficult it is to improve, i.e., the more stable the self-study completion condition. K is a function that describes this property.

[0084] R n = R o + K × (W s -W e ) (5-2)

[0085] wherein: R n is the new grade point after matching; R o is the original grade point before matching; W s is the actual self-study completion condition (1 if the self-study is completed by the subject and not completed by the opponent, 0.5 if both the subject and the opponent complete the self-study, and 0 otherwise), W e is the expected self-study completion condition, which is equal to the expected self-study completion rate; K is an uncertainty function.

[0086] For the determination of K, the following function is used:

[0087]

[0088] wherein: a, b are constant values; n is the number of matching times.

[0089] Obviously, we can analyze the self-owned data set by using the grid search method, etc. to achieve the formulation of a and b. It is worth mentioning that according to the National Bureau of Statistics Bulletin on the 2018 National Time Utilization Survey, the average learning time of citizens aged 20-24 is about 1 hour and 38 minutes. Since citizens in this age range are mainly college students, by removing the time of college students' course education, it can be known that the average self-study time of college students per day is about 40 minutes, that is, a Pomodoro. Therefore, if each learning match is a Pomodoro, then when grid searching, n should be controlled within the learning period (in days).

[0090] In addition, in order to further accelerate the convergence, we can introduce a compensation function C.

[0091] R n =R o +C×K×(W s -W e ) (5-4)

[0092]

[0093] Where: R n is the new level after matching; R o is the original level before matching; W s is the actual self-study completion (1 when the self-study is completed and the other is not, 0.5 when both are completed, and other conditions are 0), W e is the expected self-study completion, which is equal to the expected self-study completion rate; K is the uncertainty function; C is the compensation function, D is the difference between the level of the other party and oneself, and σ is the standard deviation.

[0094] The actual significance of the C function is to give double settlement points when an individual performs better than the other in a match with a self-study completion index level far superior to their own, so that their self-study completion index level converges quickly to a stable score. The matching level based on the self-study completion index level is modified as follows:

[0095] When matching users, a level is obtained for matching the current user with an appropriate object based on the self-study completion index level, that is, the matching target self-study completion index level. If the user continuously completes the self-study target, then before each match, the matching target self-study completion index level is modified by the following formula,

[0096] R m =R'+R d (6-1)

[0097]

[0098] Where: R mR' = R + (R - R) / σ * n * c c R' = R + (R - R) / σ * n * c

[0099] In other cases, the matching target self-study completion index grade point equals the self-study completion index grade point of the current user.

[0100] R m = R

[0101] R m = R

[0102] The above is only part of the specific embodiments of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. An online accompanying self-study user matching method based on a self-study completion index, characterized in that, Comprising the following steps: S1. Initialize the self-study completion index rating of each user in each learning period: For a user who newly enters the online accompanying self-study system or a user whose total number of matched sessions in the last learning period is less than ten, the self-study completion index rating of the user is initialized as: Wherein, μ is the overall average self-study completion index rating of the existing users of the online accompanying self-study system, and σ is the standard deviation; for an old user whose total number of matched sessions in the last learning period is not less than ten, the self-study completion index rating of the user is initialized by the following formula: R p = R c + D(P) where: R p is the predicted grade point for the current learning period; R c is the average grade point of all users matched in the previous period; P is the historical self-study completion rate of the current user in the previous period; D represents the difference between the two parties, the inverse function D(P) of P(D); S2. Correct the matching rating based on the self-study completion index rating, When a user is matched, a rating for matching an appropriate object for the current user is obtained by using a calculation formula based on the self-study completion index rating, i.e., a matching target self-study completion index rating; The self-study completion index rating formula is as follows: R n = R o + (W s - W e ) wherein: R n is the new rating point after matching; R o is the original rating point before matching; W s is the actual self-study completion, 1 if the self-study is completed by the own side and not completed by the other side, 0.5 if both sides complete the self-study, and otherwise; W e is the expected self-study completion, and its value is equal to the expected self-study completion rate; If a user continuously completes a self-study target, the matching target self-study completion index rating is corrected before each matching by the following formula, R m = R' + R d wherein R m is the matching target self-study completion index grade point; R' is the average self-study completion index grade point of all users in the continuous self-study completion session; R d is the matching correction score, σ is the standard deviation, n is the number of continuous self-study completion sessions in the grade point interval, and c is a constant; In other cases, the matching target self-study completion index rating is equal to the self-study completion index rating of the current user; R m = R wherein R m is the target self-study completion index grade point, and R is the current user's self-study completion index grade point; S3. Match in the rating interval in which the corrected matching target self-study completion index rating is located: In the rating interval in which the self-study completion index rating of a user is located, a user whose self-study completion index rating is closest to the self-study completion index rating of the user and whose self-study duration is the same as the self-study duration set this time is preferentially found for matching; S4. After the self-study of the matched parties is completed, the new self-study completion index ratings of the parties are calculated by using the current rating formula; R n = R o + C x K x (W s - W e ) wherein: R n is the new rating after matching; R o is the original rating before matching; W s is the actual self-study completion, 1 if the self-study is completed by the first party and not completed by the second party, 0.5 if both parties complete the self-study, and otherwise; W e is the expected self-study completion, which is equal to the expected self-study completion rate; K is an uncertainty function; C is a compensation function, D is the difference between the ratings of the other party and the self, and σ is the standard deviation; S5. Repeat steps S2-S4 several times: Whether the self-study completion index value of a user is stable is determined by using the maximum-minimum difference method.

2. The online accompanying self-study user matching method based on a self-study completion index according to claim 1, wherein, In the step S3, the specific method for preferentially finding a user whose self-study completion index rating is closest to the self-study completion index rating of the user and whose self-study duration is the same as the self-study duration set this time for matching in the rating interval in which the self-study completion index rating of the user is located is as follows: The users in the matching are divided into a plurality of matching sub-queues with a length ≤N according to the optional planned learning duration, wherein N is the maximum length of each matching sub-queue; In each sub-queue, according to the user request order, a user whose self-study completion index rating is closest to the self-study completion index rating of the user is found for matching by using the first come first served (FCFS) algorithm, and the user is matched and removed from the queue until the matching request of all users in the queue is satisfied; When a sub-queue is full, a new matching request is placed in the waiting queue corresponding to the sub-queue; when a sub-queue is not full, the matching request is inserted into the corresponding matching sub-queue in order; if a user in a sub-queue is timed out, it indicates that there is only one user in the queue, and the user is transferred to a non-empty sub-queue whose target self-study duration is closest to the target self-study duration set for the user, and the step S3 is repeated for matching; The step S3 is repeated until the matching request of all users in all queues is satisfied, at which time the request of the last user may not be satisfied, and the system prompts that the matching fails, and the matching is ended.

3. The online accompanying self-study user matching method based on a self-study completion index according to claim 1, wherein, In the step S3, the size of the hierarchical partition is: The following, where μ is the overall average self-study completion index rating of the existing users of the online self-study system, and σ is the standard deviation.

4. The online accompanying self-study user matching method based on a self-study completion index according to claim 3, wherein, In S5, the specific method for judging whether the user self-study completion index value is stable by using the maximum-minimum difference method is as follows: When the total number of fields in the current learning period of the user is greater than ten, the system will obtain the maximum value Max, the minimum value Min and the average value Avg of the user self-study completion index value from all matching records in a week and the last ten matching records, If (Max-Min) / Avg<5%, it is considered that the value tends to be stable, the loop is stopped, and the real self-study completion index of the user in this learning period is obtained, which is used for more accurate user matching.

Images