METHOD AND APPARATUS FOR PREDICTING TEST SCORES

1 . A method for predicting a test score of a user through an artificial intelligence model by a terminal, comprising:

a step of delivering training data of the user to a first layer for embedding;

a step of embedding the training data through the first layer;

a step of delivering an embedding vector from the first layer to a second layer including a compressive transformer;

a step of delivering an output value from the second layer to a third layer for predicting the test score; and

a step of outputting a prediction value for predicting the test score from the third layer.

2 . The method according to claim 1 , wherein the training data are configured with sets of pairs of questions and correct answers of the user about the questions,

wherein the embedding vector is created on the basis of the following equation:

x n =E q ( q n ) +E a ( a n−1 ), and

wherein the x n means an n-th embedding vector, the E q means an embedding layer related to the question, E a means an embedding layer related to the correct answer, q n means an n-th question, and a n−1 means a (n−1)-th correct answer.

3 . The method according to claim 2 , wherein the second layer includes an attention mask matrix, and

wherein the attention mask matrix is an upper triangular matrix.

4 . The method according to claim 3 , wherein a pre-training model for the artificial intelligence model uses, at a specific time point, only data created before the specific time point on the basis of the upper triangular matrix to perform pre-training.

5 . A terminal which predicts a test score of a user through an artificial intelligence model, comprising:

a memory which includes the artificial model; and

a processor which functionally controls the memory,

wherein the processor delivers training data of the user to a first layer for embedding, embeds the training data through the first layer, delivers an embedding vector from the first layer to a second layer including a compressive transformer, delivers an output value from the second layer to a third layer for predicting the test score, and outputs a prediction value for predicting the test score.

6 . The terminal according to claim 5 , wherein the training data are configured with sets of pairs of questions and correct answers of the user about the questions,

wherein the embedding vector is created on the basis of the following equation:

x n =E q ( q n ) +E a ( a n−1 )

wherein the x n means an n-th embedding vector, the E q means an embedding layer related to the question, E a means an embedding layer related to the correct answer, q n means an n-th question, and a n−1 means a (n−1)-th correct answer.