CN108389082B - Intelligent game rating method and system - Google Patents
Intelligent game rating method and system Download PDFInfo
- Publication number
- CN108389082B CN108389082B CN201810213868.8A CN201810213868A CN108389082B CN 108389082 B CN108389082 B CN 108389082B CN 201810213868 A CN201810213868 A CN 201810213868A CN 108389082 B CN108389082 B CN 108389082B
- Authority
- CN
- China
- Prior art keywords
- online
- game
- games
- text
- text data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0282—Rating or review of business operators or products
Abstract
The invention relates to a game intelligent rating method and a system, wherein the method comprises the following steps: collecting numerical data and text data of an online game before online; extracting numerical characteristics and text characteristics of the online game before online and performing characteristic processing; establishing and training an optimal random forest algorithm model according to the numerical characteristics, the text characteristics and the level labels of the online games before the online games are online after the characteristic processing; collecting numerical data and text data of offline games; extracting numerical characteristics and text characteristics of the offline game and performing characteristic processing; and according to the optimal random forest algorithm model, inputting the numerical characteristics and text characteristics of the offline game after characteristic processing, and predicting the game level of the offline game. The intelligent game rating method and system have the advantages of accurate rating and no influence of artificial subjective factors.
Description
Technical Field
The invention relates to the field of game rating, in particular to an intelligent game rating method and system.
Background
The gaming industry is a highly technology and content-integrated field. With the economic development, new technologies such as VR/AR and the like and the iterative updating of hardware and the reduction of Enger coefficients in the global range, the game industry is endowed with new and huge development potential and business opportunity. In the gaming industry chain, game platform operators are an important ring connecting game developers and users. The game platform operator formulates an operation strategy and distributes resources according to the game rating, so that the investment cost is in direct proportion to the division return, and the profit is maximized.
However, in the existing game rating method, an evaluator of a game platform operator generally subjectively judges a game based on his own cognition and experience, so that a game rating result is interfered by human factors such as subjective consciousness, cognitive level, experience, and staff mobility of a game evaluation group, so that the rating accuracy rate greatly fluctuates, and finally, the game platform operator is adversely affected.
Disclosure of Invention
Based on this, the present invention provides an intelligent rating method and system for a game, which have the advantages of accurate rating and no influence from human subjective factors.
A game intelligent rating method is characterized by comprising the following steps:
collecting numerical data and text data of an online game before online, wherein the numerical data is parameters of the online game, and the text data comprises short text data and long text data;
extracting numerical characteristics and text characteristics of an online game before online and performing characteristic processing, wherein the text characteristics comprise short text characteristics and long text characteristics;
establishing and training an optimal random forest algorithm model according to the numerical value characteristics and the text characteristics of the online game before the online game is online after the characteristic processing and the game level label of the online game, wherein the game level label is as follows: the game level is established according to the popularity and the profitability in a period of time after the online game is online;
collecting numerical data and text data of offline games;
extracting numerical characteristics and text characteristics of the offline game and performing characteristic processing;
and according to the optimal random forest algorithm model, inputting the numerical characteristics and text characteristics of the offline game after characteristic processing, and predicting the game level of the offline game.
A game intelligent rating system, comprising:
the online game grading module is used for finishing a game grade label of an online game based on the definition of a game grade;
the data acquisition and feature extraction module acquires numerical data and text data of online games before online and numerical data and text data of offline games, extracts corresponding numerical features and text features and performs feature processing;
the random forest training module is used for establishing and training an optimal random forest algorithm model based on numerical characteristics, text characteristics and game level labels of online games before the online games are online;
and the offline game rating module is used for predicting and rating the offline game based on the optimal random forest algorithm model and the numerical characteristics and text characteristics of the offline game.
The invention relates to an intelligent game rating method and system, which utilize numerical data, text data and game grade labels of online games before online to establish and train a random forest algorithm model; and predicting the level of the game which is not on-line based on the optimal model, the numerical data of the game which is not on-line and the text data. The method and the system are based on a game rating mechanism of machine learning, so that the game rating is not influenced by human interference factors such as subjective consciousness, cognitive level, experience abundance and personnel flow of an evaluation group, and decision support is provided for game platform operators to make reasonable operation strategies.
Further, the specific manner of the feature processing includes: performing one-hot coding on enumerated numerical characteristics or text characteristics; and carrying out discretization representation on continuous numerical characteristics or text characteristics.
Further, the step of extracting the numerical features and the text features of the online game before online comprises extracting short text features of the online game before online, and specifically comprises the following steps:
for each short text data, converting the short text data of each online game before online into a form represented by TF-IDF weight vector space, and recording a text corpus corresponding to the short text data;
classifying the online games converted into TF-IDF weight vector space forms through LDA clustering, dividing all the online games into N types of games with the same hidden theme, taking the hidden theme of each online game as the short text characteristics of the online games, and recording N clustering centers corresponding to the short text data, wherein N is a constant.
Further, the step of extracting the numerical features and the text features of the online game before online includes extracting long text features of the online game before online, where the long text features include long text features without emotion factors and long text features with emotion factors, and for the long text features without emotion factors, the step of extracting the long text features of the online game before online specifically includes the following steps:
for each kind of long text data without emotional factors, the word segmentation tool is utilized to cut the long text data without the emotional factors of each online game before online into a plurality of short text data;
combining the obtained short text data in pairs, converting the short text data into a TF-IDF weight vector space form, and simultaneously recording a text corpus corresponding to the long text data without emotion factors;
classifying the online games converted into TF-IDF weight vector space forms through LDA clustering, dividing all the online games into N types of games with the same hidden theme, taking the hidden theme to which each online game belongs as the long text features of the online games without emotional factors, and recording N clustering centers corresponding to the long text data of the online games without emotional factors, wherein N is a constant.
Further, for the long text feature with the emotional factors, the extracting the long text feature before the online game is online specifically includes the following steps:
establishing a positive word stock, a negative word stock, a degree word stock and a negative word stock, and specifying the emotion scores of each word in the positive word stock, the negative word stock, the degree word stock and the negative word stock;
configuring a positive word stock, a negative word stock, a degree word stock and a negative word stock into a corpus;
for each long text data with emotional factors, the word segmentation tool is utilized to cut each long text data with emotional factors before online game is online into a plurality of short text data;
traversing short text data, checking whether a positive word or a negative word exists, if not, the emotion score is zero, if so, checking whether the previous word is a degree word or a negative word, if so, continuously checking whether the previous word is the degree word or the negative word until the detected previous word is not the degree word or the negative word, simultaneously checking whether the next word is the degree word or the negative word, if so, continuously checking whether the next word is the degree word or the negative word until the detected next word is not the degree word or the negative word, wherein the emotion score is equal to the sum of the emotion scores of each existing positive word or negative word multiplied by the emotion scores of all corresponding prefixes and suffix degree words or negative words, and the emotion score of each online game is used as the long text characteristic of the online game with emotion factors.
Further, the step of extracting the numerical features and the text features of the offline game includes extracting short text features of the offline game, and specifically includes the following steps:
for each short text data, according to the recorded text corpus corresponding to the short text data in the specific step of extracting the short text features of the online games before the online games are online, converting the short text data of each offline game into a form expressed by a TF-IDF weight vector space;
according to the specific step of extracting the short text features of the online games before online, classifying the offline games converted into TF-IDF weight vector space form by the recorded N clustering centers corresponding to the short text data, dividing all the offline games into N types of games with the same hidden theme, and taking the hidden theme of each offline game as the short text features of the offline games.
Further, the step of extracting the numerical features and the text features of the offline game includes extracting long text features of the offline game, where the long text features include long text features without emotion factors and long text features with emotion factors, and for the long text features without emotion factors, the step of extracting the long text features of the offline game specifically includes the following steps:
for each kind of long text data without emotional factors, the word segmentation tool is utilized to cut the long text data without emotional factors of each offline game into a plurality of short text data;
combining the obtained short text data in pairs, and converting the recorded text corpus corresponding to the long text data without the emotional factors into TF-IDF weight vector space representation according to the specific step of extracting the long text features without the emotional factors before the online game is online;
according to the specific step of extracting the long text features without the emotion factors before online of the online game, the recorded N clustering centers corresponding to the long text data without the emotion factors classify the offline games converted into TF-IDF weight vector space forms, divide all the offline games into N types of games with the same hidden theme, and take the hidden theme of each offline game as the long text features without the emotion factors of the offline games.
A computer readable medium storing a computer program which when executed by a processor implements the steps of any of the above game intelligent rating methods.
A computer device comprising a memory, a processor and a computer program stored in said memory and executable by said processor, said processor implementing the steps of the game intelligence rating method of any of the above when executing said computer program.
For a better understanding and practice, the invention is described in detail below with reference to the accompanying drawings.
Drawings
FIG. 1 is a flow chart of a method for intelligent rating of games in an embodiment of the present invention;
FIG. 2 is a flowchart illustrating the extraction of short text features of an online game before online according to an embodiment of the present invention;
FIG. 3 is a flowchart of extracting features of a long text without emotion factors before an online game is online according to an embodiment of the present invention;
FIG. 4 is a flowchart illustrating the steps of extracting long text features with emotion factors before online game play according to an embodiment of the present invention;
FIG. 5 is a flowchart of extracting short text features of an offline game according to an embodiment of the present invention;
FIG. 6 is a flowchart of extracting long text features without emotion factors for a non-online game according to an embodiment of the present invention;
FIG. 7 is a schematic diagram of a game intelligence rating system in an embodiment of the present invention.
Detailed Description
As shown in FIG. 1, in one embodiment of the invention, a game intelligence rating method of the invention comprises the following steps:
s11 collecting numerical data and text data of the online game before online, wherein the numerical data is parameters of the online game, and the text data comprises short text data and long text data;
in this embodiment, the numerical data includes whether to play a stand-alone game, a game selectable language number, a game mix situation, whether to have a novice guide, a game fluency, a game loading time, a game screen refinement degree, a game scene type, and the like.
The short text data includes game features, game play, game types, and game themes. The game features may include 3D, animation, quadratic, battle, little fresh, etc.; game play may include FPS, TPS, MMO, MMOARPG, MOBA, etc.; game types may include policy, action, flight, fighting, character, etc.; the game subject matter may include oriental hallucinations, ancient war, domestic animation, Korean fantasy, European and American fantasy, etc.
The long text data includes game descriptions and game fan commentary.
S12, extracting the numerical characteristic and the text characteristic of the online game before the online game and carrying out characteristic processing, wherein the text characteristic comprises a short text characteristic and a long text characteristic.
In this embodiment, the specific manner of the feature processing includes: performing one-hot coding on enumerated numerical characteristics or text characteristics; carrying out discretization representation on continuous numerical characteristics or text characteristics;
as shown in fig. 2, the method for extracting the short text feature before online game specifically includes the following steps:
s21, for each short text data, converting the short text data before online of each online game into a form expressed by TF-IDF weight vector space, and recording a text corpus corresponding to the short text data;
the text data is converted into TF-IDF weight vector space representation by a method which is characterized in that different text data of all samples are integrated into a vocabulary V, each sample text data corresponds to a V-dimensional vector, and elements in the vector take TF-IDF values corresponding to participles; the normalized vector is the TF-IDF weight vector space representation corresponding to the sample text data.
Wherein, the calculation method of the word segmentation TF-IDF value is TF-IDF. TF is the number of times that the word segmentation appears in the sample text data; IDF represents the logarithm of the frequency of occurrence of the word segmentation in all sample text data, i.e.:
n is the number of samples and DF is the number of samples containing the participle.
The specific calculation method of the normalization vector comprises the following steps:
s22 classifies each online game converted into TF-IDF weight vector space form by LDA clustering, divides all online games into N types of games with the same hidden theme, takes the hidden theme of each online game as the short text feature of the online game, and records N clustering centers corresponding to the short text data, wherein N is a constant.
In this embodiment, N may take 3.
The long text features include long text features without emotion factors and long text features with emotion factors, and for the long text features without emotion factors, as shown in fig. 3, the extracting long text features of the online game before online specifically includes the following steps:
s31, for each long text data without emotion factors, cutting each long text data without emotion factors before online game into a plurality of short text data by using a word segmentation tool;
s32, carrying out adjacent pairwise combination on the obtained short text data, converting the short text data into TF-IDF weight vector space representation, and recording the text corpus corresponding to the long text data without emotion factors;
s33 classifies each online game converted into TF-IDF weight vector space form through LDA clustering, divides all online games into N types of games with the same hidden theme, takes the hidden theme of each online game as the long text feature without emotion factors of the online game, and records N clustering centers corresponding to the long text data without emotion factors of the online game, wherein N is a constant.
In this embodiment, N may take 3.
As shown in fig. 4, for the long text feature with emotional factors, the extracting the long text feature before the online game is online specifically includes the following steps:
s41, establishing a positive lexicon, a negative lexicon, a degree lexicon and a negative lexicon, and designating the emotion score of each word in the positive lexicon, the negative lexicon, the degree lexicon and the negative lexicon;
s42, configuring the positive word stock, the negative word stock, the degree word stock and the negative word stock into a corpus;
s43, for each long text data with emotional factors, cutting each long text data with emotional factors before online game into a plurality of short text data by using a word segmentation tool;
specifically, in this embodiment, the word segmentation tool may select jieba word segmentation software.
S44, traversing the short text data, checking whether positive words or negative words exist, if not, the emotion score is zero, if so, checking whether the previous word is a degree word or a negative word, if so, continuously checking whether the previous word is a degree word or a negative word until the detected previous word is not a degree word or a negative word, simultaneously checking whether the next word is a degree word or a negative word, if so, continuously checking whether the next word is a degree word or a negative word until the detected next word is not a degree word or a negative word, wherein the emotion score is equal to the sum of the emotion score of each existing positive word or negative word multiplied by the emotion scores of all prefixes and suffix degree words or negative words, and the emotion score of each online game is used as the long text characteristic of the online game with emotion factors.
S13, establishing and training an optimal random forest algorithm model according to the numerical value characteristics and the text characteristics of the online game before the online game is processed by the characteristics and the game level labels of the online game, wherein the game level labels are as follows: the game level is established according to the popularity and the profitability in a period of time after the online game is online;
specifically, the method for establishing and training the optimal random forest algorithm model comprises the following steps: establishing a random forest algorithm model based on numerical data, text characteristics and game level labels of online games before online; and then setting an assessment index, and training by a K-fold cross validation method to obtain an optimal game intelligent rating model.
Specifically, the random forest algorithm model is, for sample xi∈RnI 1, …, and a corresponding label y e RlAnd establishing a plurality of decision trees. Each decision tree is initially put back
And (4) sampling. Based on greedy algorithm, the optimal space is divided recursively, and each time the space is divided, the space is taken
Until a termination condition is not partitionable or triggered. Pruning each decision tree based on a penalty rule, and taking an average value of results of a plurality of decision trees as a final label.
The method for dividing the optimal space is specifically that data at a node m is assumed to be Q, and a dividing behavior θ is (j, t)m) According to the characteristic j according to the threshold tmDivide, then data Q is divided into Qleft(theta) and Qright(theta), respectively calculating Qleft(theta) and QrightCross entropy of (θ) H (Q)left(theta)) and H (Q)right(theta)), the optimal division behavior theta*Comprises the following steps:
θ*＝argminθ(nleftH(Qleft(θ))+nrightH(Qright(θ)))
wherein, the cross entropy calculation method is specifically to assume the label yiIt may be 0,1, …, K-1 in the presence of NmData of node m of one sample is Q, yiThe probability of belonging to a k-tag is:
pmk＝1/Nm∑I(yi＝k)，
then the cross entropy of data Q is:
H(Q)＝-∑kpmklog(pmk)。
in the present embodiment, the assessment indexes may include logarithmic loss, precision, recall, F1-Score, and the like. K may take 10.
S14 collecting numerical data and text data of the offline game;
s15 extracting the numerical characteristic and text characteristic of the offline game and performing characteristic processing;
as shown in fig. 5, the extracting of the short text feature of the offline game specifically includes the following steps:
s51, converting the short text data of each off-line game into a form expressed by TF-IDF weight vector space according to the recorded text corpus corresponding to the short text data in the step S21 for each kind of short text data;
s52, according to the recorded N clustering centers corresponding to the short text data in the step S22, classifying the games which are not online and are converted into TF-IDF weight vector space form, dividing all the games which are not online into N types of games with the same implied subject, and taking the implied subject of each game which is not online as the short text feature of the games which are not online.
As shown in fig. 6, the extracting long text features without emotion factors of the offline game specifically includes the following steps:
s61, for each long text data without emotion factors, cutting the long text data without emotion factors of each offline game into a plurality of short text data by using a word segmentation tool;
s62, combining the obtained short text data two by two, and converting the short text data into TF-IDF weight vector space representation according to the recorded text corpus corresponding to the long text data without emotion factors in the step S32;
s63, according to the recorded N clustering centers corresponding to the long text data without emotion factors in the step S33, classifying the games which are not online and converted into TF-IDF weight vector space form, dividing all the games which are not online into N types of games with the same hidden theme, and taking the hidden theme of each game which is not online as the long text features without emotion factors of the games which are not online.
S16, according to the optimal random forest algorithm model, inputting the numerical characteristics and text characteristics of the offline game after characteristic processing, and predicting the game level of the offline game.
And inputting the optimal random forest algorithm model by taking the numerical characteristics and the text characteristics of the offline game as input variables, and outputting the game level label of the offline game by the optimal random forest algorithm model through calculating the input variables, thereby finishing the rating of the offline game.
In another embodiment of the present invention, as shown in FIG. 7, the present invention further comprises a game intelligence rating system comprising:
the online game grading module 10 is used for finishing a game grade label of an online game based on the definition of a game grade;
and the data acquisition and feature extraction module 20 is used for acquiring numerical data and text data of online games before online, and numerical data and text data of offline games, extracting corresponding numerical features and text features, and performing feature processing.
A random forest training module 30, which establishes and trains an optimal random forest algorithm model based on the numerical characteristics, text characteristics and game level labels of the online game before online
And the offline game rating module 40 is used for predicting and rating the offline game based on the optimal random forest algorithm model and the numerical characteristics and text characteristics of the offline game.
In another embodiment of the present invention, a computer readable medium is further included, which stores a computer program, and the computer program is executed by a processor to implement the steps of the game intelligence rating method and system in any one of the above embodiments.
In another embodiment of the present invention, a computer device is further included, which includes a memory, a processor, and a computer program stored in the storage and executable by the processor, and the processor executes the computer program to implement the steps of the game intelligence rating method and system in any of the above embodiments.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention.
Claims (8)
1. A game intelligent rating method is characterized by comprising the following steps:
collecting numerical data and text data of an online game before online, wherein the numerical data is parameters of the online game, and the text data comprises short text data and long text data;
extracting numerical characteristics and text characteristics of an online game before online and performing characteristic processing, wherein the text characteristics comprise short text characteristics and long text characteristics;
establishing and training an optimal random forest algorithm model according to the numerical value characteristics and the text characteristics of the online game before the online game is online after the characteristic processing and the game level label of the online game, wherein the game level label is as follows: the game level is established according to the popularity and the profitability in a period of time after the online game is online;
collecting numerical data and text data of offline games;
extracting numerical characteristics and text characteristics of the offline game and performing characteristic processing;
according to the optimal random forest algorithm model, inputting numerical characteristics and text characteristics of the offline game after characteristic processing, and predicting the game level of the offline game;
wherein, the step of extracting the numerical characteristic and the text characteristic of the online game before the online game comprises the step of extracting the short text characteristic of the online game before the online game,
for each short text data, converting the short text data of each online game before online into a form represented by TF-IDF weight vector space, and recording a text corpus corresponding to the short text data;
classifying the online games converted into TF-IDF weight vector space forms through LDA clustering, dividing all the online games into N types of games with the same hidden theme, taking the hidden theme to which each online game belongs as the short text characteristics of the online games, and recording N clustering centers corresponding to the short text data, wherein N is a constant;
the step of extracting the numerical characteristics and the text characteristics of the offline game comprises the step of extracting the short text characteristics of the offline game, in particular comprising the following steps,
for each short text data, converting the short text data of each offline game into a form represented by a TF-IDF weight vector space according to the recorded text corpus corresponding to the short text data;
and classifying the games which are not online and converted into TF-IDF weight vector space forms according to the recorded N clustering centers corresponding to the short text data, dividing all the games which are not online into N types of games with the same hidden theme, and taking the hidden theme to which each game which is not online belongs as the short text feature of the game.
2. The intelligent rating method for games of claim 1, wherein the specific way of processing the features comprises performing one-hot coding on enumerated numerical features or text features, and performing discretization representation on continuous numerical features or text features.
3. The intelligent rating method for games as claimed in claim 1, wherein the step of extracting the numerical features and the text features of the online game before online comprises extracting the long text features of the online game before online, wherein the long text features comprise the long text features without emotion factors and the long text features with emotion factors, and the step of extracting the long text features of the online game before online comprises the following steps:
for each kind of long text data without emotional factors, the word segmentation tool is utilized to cut the long text data without the emotional factors of each online game before online into a plurality of short text data;
combining the obtained short text data in pairs, converting the short text data into a TF-IDF weight vector space form, and recording a text corpus corresponding to the long text data without emotional factors;
classifying the online games converted into TF-IDF weight vector space forms through LDA clustering, dividing all the online games into N types of games with the same hidden theme, taking the hidden theme to which each online game belongs as the long text features of the online games without the emotional factors, and recording N clustering centers corresponding to the long text data of the online games without the emotional factors, wherein N is a constant.
4. The intelligent rating method for games according to claim 3, wherein for the long text features with emotional factors, the extracting the long text features before online game specifically comprises the following steps:
establishing a positive word stock, a negative word stock, a degree word stock and a negative word stock, and specifying the emotion scores of each word in the positive word stock, the negative word stock, the degree word stock and the negative word stock;
configuring a positive word stock, a negative word stock, a degree word stock and a negative word stock into a corpus;
for each long text data with emotional factors, the word segmentation tool is utilized to cut each long text data with emotional factors before online game is online into a plurality of short text data;
traversing short text data, checking whether a positive word or a negative word exists, if not, the emotion score is zero, if so, checking whether the previous word is a degree word or a negative word, if so, continuously checking whether the previous word is the degree word or the negative word until the detected previous word is not the degree word or the negative word, simultaneously checking whether the next word is the degree word or the negative word, if so, continuously checking whether the next word is the degree word or the negative word until the detected next word is not the degree word or the negative word, wherein the emotion score is equal to the sum of the emotion scores of each existing positive word or negative word multiplied by the emotion scores of all corresponding prefixes and suffix degree words or negative words, and the emotion score of each online game is used as the long text characteristic of the online game with emotion factors.
5. The intelligent rating method for games of claim 3, wherein the step of extracting the numerical features and the text features of the offline game comprises extracting long text features of the offline game, wherein the long text features comprise long text features without emotion factors and long text features with emotion factors, and the step of extracting the long text features of the offline game specifically comprises the following steps:
for each long text data without emotional factors, cutting the long text data of each offline game into a plurality of short text data by using a word segmentation tool;
combining the obtained short text data in pairs, and converting the short text data into TF-IDF weight vector space representation according to the recorded text corpus corresponding to the long text data without emotional factors;
and classifying the offline games converted into TF-IDF weight vector space forms according to the recorded N clustering centers corresponding to the long text data without the emotional factors, dividing all the offline games into N types of games with the same hidden theme, and taking the hidden theme to which each offline game belongs as the long text features without the emotional factors of the offline games.
6. A game intelligent rating system, comprising:
the online game grading module is used for finishing a game grade label of an online game based on the definition of a game grade;
the data acquisition and feature extraction module acquires numerical data and text data of online games before online and numerical data and text data of offline games, extracts corresponding numerical features and text features and performs feature processing;
the random forest training module is used for establishing and training an optimal random forest algorithm model based on numerical characteristics, text characteristics and game level labels of online games before the online games are online;
the offline game rating module is used for predicting and rating the offline game based on the optimal random forest algorithm model and the numerical characteristics and text characteristics of the offline game;
wherein, the step of extracting the numerical characteristic and the text characteristic of the online game before the online game comprises the step of extracting the short text characteristic of the online game before the online game,
for each short text data, converting the short text data of each online game before online into a form represented by TF-IDF weight vector space, and recording a text corpus corresponding to the short text data;
classifying the online games converted into TF-IDF weight vector space forms through LDA clustering, dividing all the online games into N types of games with the same hidden theme, taking the hidden theme to which each online game belongs as the short text characteristics of the online games, and recording N clustering centers corresponding to the short text data, wherein N is a constant;
the step of extracting the numerical characteristics and the text characteristics of the offline game comprises the step of extracting the short text characteristics of the offline game, in particular comprising the following steps,
for each short text data, converting the short text data of each offline game into a form represented by a TF-IDF weight vector space according to the recorded text corpus corresponding to the short text data;
and classifying the games which are not online and converted into TF-IDF weight vector space forms according to the recorded N clustering centers corresponding to the short text data, dividing all the games which are not online into N types of games with the same hidden theme, and taking the hidden theme to which each game which is not online belongs as the short text feature of the games which are not online.
7. A computer-readable medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the game intelligent rating method according to any one of claims 1 to 5.
8. A computer device comprising a memory, a processor and a computer program stored in said memory and executable by said processor, said processor implementing the steps of the game intelligence rating method of any one of claims 1 to 5 when executing said computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810213868.8A CN108389082B (en) | 2018-03-15 | 2018-03-15 | Intelligent game rating method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810213868.8A CN108389082B (en) | 2018-03-15 | 2018-03-15 | Intelligent game rating method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108389082A CN108389082A (en) | 2018-08-10 |
| CN108389082B true CN108389082B (en) | 2021-07-06 |
Family
ID=63067731
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810213868.8A Active CN108389082B (en) | 2018-03-15 | 2018-03-15 | Intelligent game rating method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108389082B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107193801A (en) * | 2017-05-21 | 2017-09-22 | 北京工业大学 | A kind of short text characteristic optimization and sentiment analysis method based on depth belief network |
| CN107315797A (en) * | 2017-06-19 | 2017-11-03 | 江西洪都航空工业集团有限责任公司 | A kind of Internet news is obtained and text emotion forecasting system |
| CN107506461A (en) * | 2017-08-30 | 2017-12-22 | 北京非斗数据科技发展有限公司 | A kind of method of intellectual analysis and risk assessment on video display drama |
| CN110797046A (en) * | 2018-08-02 | 2020-02-14 | 中国移动通信集团广东有限公司 | Method and device for establishing prediction model of voice quality MOS value |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9792534B2 (en) * | 2016-01-13 | 2017-10-17 | Adobe Systems Incorporated | Semantic natural language vector space |
-
2018
- 2018-03-15 CN CN201810213868.8A patent/CN108389082B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107193801A (en) * | 2017-05-21 | 2017-09-22 | 北京工业大学 | A kind of short text characteristic optimization and sentiment analysis method based on depth belief network |
| CN107315797A (en) * | 2017-06-19 | 2017-11-03 | 江西洪都航空工业集团有限责任公司 | A kind of Internet news is obtained and text emotion forecasting system |
| CN107506461A (en) * | 2017-08-30 | 2017-12-22 | 北京非斗数据科技发展有限公司 | A kind of method of intellectual analysis and risk assessment on video display drama |
| CN110797046A (en) * | 2018-08-02 | 2020-02-14 | 中国移动通信集团广东有限公司 | Method and device for establishing prediction model of voice quality MOS value |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108389082A (en) | 2018-08-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Memmert et al. | Match analysis, big data and tactics: current trends in elite soccer. | |
| Hossain et al. | " President Vows to Cut< Taxes> Hair": Dataset and Analysis of Creative Text Editing for Humorous Headlines | |
| Santos et al. | Boosting named entity recognition with neural character embeddings | |
| CN108304526B (en) | Data processing method and device and server | |
| CN108304853B (en) | Game correlation obtaining method and device, storage medium and electronic device | |
| EP4125029A1 (en) | Electronic apparatus, controlling method of thereof and non-transitory computer readable recording medium | |
| Canossa | Meaning in gameplay: Filtering variables, defining metrics, extracting features and creating models for gameplay analysis | |
| CN113392331A (en) | Text processing method and equipment | |
| CN108717459B (en) | A kind of mobile application defect positioning method of user oriented comment information | |
| CN111125360A (en) | Emotion analysis method and device in game field and model training method and device thereof | |
| US20150269162A1 (en) | Information processing device, information processing method, and computer program product | |
| Rodriguez-Fernandez et al. | A study on performance metrics and clustering methods for analyzing behavior in UAV operations | |
| CN115463424A (en) | Display control method and device of virtual role and electronic equipment | |
| CN115662627A (en) | Early warning method for building early warning model of psychological health risks of students and application of early warning method | |
| CN113617036A (en) | Game dialogue processing method, device, equipment and storage medium | |
| CN112245934B (en) | Data analysis method, device and equipment for virtual resources in virtual scene application | |
| CN108389082B (en) | Intelligent game rating method and system | |
| Kummer et al. | Applying commitment to churn and remaining players lifetime prediction | |
| CN112819499A (en) | Information transmission method, information transmission device, server and storage medium | |
| CN110263344B (en) | Text emotion analysis method, device and equipment based on hybrid model | |
| Xie et al. | Predicting disengagement in free-to-play games with highly biased data | |
| CN111309873A (en) | Data processing method and device, electronic equipment and storage medium | |
| CN107832304A (en) | A kind of method and system that user's sex is judged based on Message-text | |
| CN113569018A (en) | Question and answer pair mining method and device | |
| Indulkar | PUBG Winner Ranking Prediction using R Interface ‘h2o’Scalable Machine Learning Platform |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |