A crime prediction method and system based on formal concept analysis

Abstract

The present invention provides a crime prediction method and system based on formal concept analysis. The method includes: step 1: processing the existing case judgment documents and relevant laws and regulations based on natural language processing technology to obtain the criminal acts and crimes in the existing case judgment documents and relevant laws and regulations; step 2: processing the criminal acts and crimes in the existing case judgment documents and relevant laws and regulations into formal backgrounds, and using formal concept analysis methods to process the formal backgrounds to generate a series of value-dependent implications corresponding to criminal acts and crimes; step 3: processing the case to be predicted based on natural language processing technology to obtain the criminal acts contained in the case to be predicted, matching the criminal acts in the case to be predicted with a series of value-dependent implications, and obtaining the crimes corresponding to the criminal acts in the case to be predicted.

Description

Technical Field

[0001] The present invention relates to the field of data mining technology, and in particular to a crime prediction method and system based on formal concept analysis. Background Art

[0002] Nowadays, the number of court case judgments has increased significantly, but the number of relevant practitioners is not enough to handle a large number of cases in a timely manner. Therefore, how to quickly filter out key information from case judgments and predict the crime verdict of the case based on the key information and relevant laws and regulations is a current research hotspot in crime prediction.

[0003] Most current methods use a method based on data mining and deep learning to predict the charges of most cases. This method can provide possible charges for reference by relevant staff. However, there are still some problems with the existing methods. The reason is that most of the existing methods use models constructed based on existing cases to predict unprocessed cases. This method has a high accuracy rate when facing common cases, but when facing rare cases, the prediction results often have large deviations. In addition, many cases in real life contain multiple charges, but most of the current charge prediction methods can only predict one charge. Therefore, when facing multiple charges at the same time, the existing methods cannot play a good effect. Summary of the invention

[0004] In view of the problem that traditional crime prediction methods have low prediction accuracy or cannot predict multiple crimes at the same time when facing relatively rare cases, the present invention provides a crime prediction method and system based on formal concept analysis.

[0005] On the one hand, the present invention provides a crime prediction method based on formal concept analysis, comprising:

[0006] Step 1: Process the existing case judgment documents and relevant laws and regulations based on natural language processing technology to obtain the criminal acts and charges in the existing case judgment documents and relevant laws and regulations;

[0007] Step 2: Treat the criminal acts and charges in the existing case judgment documents and relevant laws and regulations as formal background, and use the formal concept analysis method to process the formal background to generate a series of value-dependent implication formulas corresponding to the criminal acts and charges;

[0008] Step 3: Process the case to be predicted based on natural language processing technology to obtain the criminal acts contained in the case to be predicted, match the criminal acts in the case to be predicted with a series of value-dependent implications, and obtain the crime corresponding to the criminal acts in the case to be predicted.

[0009] Furthermore, in step 2, the criminal acts and charges in the existing case judgment documents and relevant laws and regulations are processed as formal background, specifically including:

[0010] Step A1: Take the criminal act and crime as attributes in the formal background, and take each existing case judgment document or relevant laws and regulations as an object in the formal background;

[0011] Step A2: For each object, if the object has the corresponding attribute, mark the intersection cell of the object and the attribute as 1, otherwise mark it as 0, until all objects are marked, and obtain a table containing only 0, 1, object, and attribute. This table is the formal background.

[0012] Furthermore, in step 2, the formal concept analysis method is used to process the formal background to generate a series of value-dependent implication formulas corresponding to criminal acts and crimes, specifically including:

[0013] Step B1: extract all attributes in the formal background and sort all attributes in lexicographic order;

[0014] Step B2: for each attribute set in the lexicographic order, verify whether B→f(g(B))-B holds in order from the smallest to the largest lexicographic order, where B is the attribute set in the lexicographic order; if B→f(g(B))-B holds, then continue to determine whether B is equal to f(g(B)); if B is equal to f(g(B)), then add B to the concept connotation set; if B is not equal to f(g(B)), then add B→f(g(B))-B to the value-dependent implication set;

[0015] Step B3: After processing all attribute sets in the lexicographic order according to step B2, all value-dependent implication formulas in the value-dependent implication formula set are screened, and the screening condition is: deleting the value-dependent implication formulas whose value-dependent front key contains a decision attribute or whose value-dependent rear key contains a conditional attribute;

[0016] Step B4: For all the value-dependent implication formulas after screening, the attributes in the value-dependent implication formulas are matched with the crime names and / or criminal acts, that is, a series of value-dependent implication formulas corresponding to the criminal acts and the crime names are obtained.

[0017] Furthermore, in step B1, the step of sorting all attributes in lexicographic order specifically includes:

[0018] Step B11: Define the order of the attributes. Specifically, if all criminal acts are classified as a1, a2, a3...a n1 It means that all the crimes are classified into b1, b2, b3...bn2 Indicates that the order of the size of all attributes is defined as: a1<a2<a3<...<a n1 <b1<b2<b3<...<b n2 ;

[0019] Step B12: define the size order between attribute sets, specifically: set the set of all attributes G = {a1, a2, a3...a n1 ,b1,b2,b3...b n2}, there is a property set and attribute collection The elements in the attribute sets A and B are sorted according to the definition in step B11. If the first unequal element in A and B from the left is larger than the first unequal element in B, then A is defined as less than B.

[0020] Step B13: sort all attribute sets according to the definition in step B12, and the obtained order from small to large is the lexicographic order; wherein each attribute set is different and satisfies the order from small to large in each attribute set.

[0021] In another aspect, the present invention provides a crime prediction system based on formal concept analysis, comprising:

[0022] The text processing module is used to process the existing case judgment documents and relevant laws and regulations based on natural language processing technology to obtain the criminal acts and charges in the existing case judgment documents and relevant laws and regulations;

[0023] The data analysis module is used to process the criminal acts and charges in the existing case judgment documents and relevant laws and regulations as formal backgrounds, and use formal concept analysis methods to process the formal backgrounds to generate a series of value-dependent implication formulas corresponding to the criminal acts and charges;

[0024] The crime prediction module is used to process the case to be predicted based on natural language processing technology, obtain the criminal behavior contained in the case to be predicted, match the criminal behavior in the case to be predicted with a series of value-dependent implications, and obtain the crime corresponding to the criminal behavior in the case to be predicted.

[0025] Further, the data analysis module includes an attribute and object extraction unit and a form background generation unit;

[0026] The attribute and object extraction unit is used to take the criminal behavior and crime as attributes in the formal background, and take each existing case judgment document or relevant laws and regulations as an object in the formal background;

[0027] The formal background generating unit is used to mark the intersection cell of each object and the attribute as 1 when the object contains the corresponding attribute, otherwise it is marked as 0, until all objects are marked, and a table containing only 0, 1, object, and attribute is obtained, and the table is used as the formal background.

[0028] Furthermore, the data analysis module includes an attribute sorting unit, an attribute verification unit, a screening unit, and a value-dependent implication generation unit;

[0029] an attribute sorting unit, for extracting all attributes in the formal background and sorting all attributes in lexicographic order;

[0030] The attribute verification unit is used to verify whether B→f(g(B))-B holds for each attribute set in the lexicographic order in ascending order of the lexicographic order, where B is the attribute set in the lexicographic order; if B→f(g(B))-B holds, then continue to judge whether B is equal to f(g(B)); if B is equal to f(g(B)), then add B to the concept connotation set; if B is not equal to f(g(B)), then add B→f(g(B))-B to the value-dependent implication set;

[0031] A screening unit is used to screen all value-dependent implication formulas in the value-dependent implication formula set obtained by the attribute verification unit after the attribute verification unit processes all attribute sets in the lexicographic order, wherein the screening condition is: deleting the value-dependent implication formulas whose value-dependent front key contains a decision attribute or whose value-dependent rear key contains a conditional attribute;

[0032] The value-dependent implication generating unit is used to correspond the attributes in the value-dependent implication with the crime and / or criminal behavior for all the value-dependent implication after screening, so as to obtain a series of value-dependent implication corresponding to the crime.

[0033] Furthermore, the attribute sorting unit is specifically used to:

[0034] Define the order of magnitude between attributes. Specifically, if all criminal acts are assigned the order of a1, a2, a3...a n1 It means that all the crimes are classified into b1, b2, b3...b n2 Indicates that the order of the size of all attributes is defined as: a1<a2<a3<...<a n1 <b1<b2<b3<...<b n2 ;

[0035] Define the size order between attribute sets, specifically: set the set of all attributes G = {a1, a2, a3...a n1 ,b1,b2,b3...bn2}, there is a property set and attribute collection The elements in the attribute sets A and B are sorted according to the definition in step B11. If the first unequal element in A and B from the left is larger than the first unequal element in B, then A is defined as less than B.

[0036] All attribute sets are sorted according to the definition, and the resulting order from small to large is the lexicographical order; wherein each attribute set is different and satisfies the order from small to large in each attribute set.

[0037] Beneficial effects of the present invention:

[0038] The crime prediction method and system based on situation concept analysis provided by the present invention introduce the original text of relevant laws and regulations when processing case judgment documents, which can effectively improve the accuracy of crime prediction when facing uncommon cases; compared with the prediction results of most current crime prediction methods, which can often only obtain a single crime, the value-dependent implication formula generated by the situation concept analysis of the present invention can be used for multiple crime predictions, which has great practical significance in assisting relevant practitioners. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic diagram of a flow chart of a crime prediction method based on situation concept analysis provided by an embodiment of the present invention;

[0040] Figure 2 A schematic diagram of the structure of a crime prediction system based on situation concept analysis provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution in the embodiment of the present invention will be clearly described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] Situational concept analysis (FCA) is a tool for data description and knowledge discovery. The present invention creatively applies situational concept analysis as a tool to crime verdict prediction, which can better discover the relationship between crime and criminal behavior, and predict crimes for cases that have not yet been judged faster and more accurately, significantly reducing the workload of relevant practitioners; and, using the value-dependent implication in situational concept analysis can significantly improve the interpretability of crime prediction.

[0043] Example 1

[0044] like Figure 1 As shown, an embodiment of the present invention provides a crime prediction method based on formal concept analysis, comprising:

[0045] S101: Process the existing case judgment documents and relevant laws and regulations based on natural language processing technology to obtain the criminal acts and charges in the existing case judgment documents and relevant laws and regulations;

[0046] Specifically, based on machine learning methods, supervised learning methods can be used to extract criminal acts and charges from existing judicial documents and relevant laws and regulations. It should be noted that after extraction, the text of criminal acts and charges should be screened to remove repeated criminal acts and charges.

[0047] S102: Treat the criminal acts and charges in the existing case judgment documents and relevant laws and regulations as formal background, and use the formal concept analysis method to process the formal background to generate a series of value-dependent implication formulas corresponding to the criminal acts and charges;

[0048] As a feasible implementation method, the criminal acts and charges in the existing case judgment documents and relevant laws and regulations are treated as formal background, specifically including:

[0049] Step A1: Take the criminal act and crime as attributes in the formal background, and take each existing case judgment document or relevant laws and regulations as an object in the formal background;

[0050] Step A2: For each object, if the object has the corresponding attribute, mark the intersection cell of the object and the attribute as 1, otherwise mark it as 0, until all objects are marked, and obtain a table containing only 0, 1, object, and attribute. This table is the formal background.

[0051] For example, attribute a1 is defined as: stolen, plundered, robbed, and stolen, that is, documents containing these words are considered to contain attribute a1; attribute a2 is defined as: false invoice, purchase, purchased, and forged, that is, documents containing these words are considered to contain attribute a2; attribute a3 is defined as invoice and invoice fee, that is, documents containing these words are considered to contain attribute a3; attribute b1 is defined as theft, that is, documents containing this word are considered to contain attribute b1; attribute b2 is defined as false invoice, that is, documents containing this word are considered to contain attribute b2.

[0052] Assuming that object 1 has attributes a1 and b1, and object 2 has attributes a2, a3, and b2, the formal background composed of object 1, object 2 and attributes a1, a2, a3, b1, and b2 is shown in Table 1.

[0053] Table 1

[0054]

[0055] As an implementable method, the formal concept analysis method is used to process the formal background to generate a series of value-dependent implication formulas corresponding to criminal acts and crimes, specifically including:

[0056] Step B1: extract all attributes in the formal background and sort all attributes in lexicographic order;

[0057] Specifically, this step includes the following sub-steps:

[0058] Step B11: Define the order of the attributes. Specifically, if all criminal acts are classified as a1, a2, a3...a n1 It means that all the crimes are classified into b1, b2, b3...b n2 Indicates that the order of the size of all attributes is defined as: a1<a2<a3<...<a n1 <b1<b2<b3<...<b n2 ;

[0059] Step B12: define the size order between attribute sets, specifically: set the set of all attributes G = {a1, a2, a3...a n1 ,b1,b2,b3...b n2}, there is a property set and attribute collection The elements in the attribute sets A and B are sorted according to the definition in step B11. If the first unequal element in A and B from the left is larger than the first unequal element in B, then A is defined as less than B.

[0060] Step B13: Sort all attribute sets according to the definition in step B12, and the resulting ascending order is the lexicographic order; wherein each attribute set is different and satisfies the ascending order in their respective attribute sets, for example: define subset A = {a1a2}, not A = {a2a1}.

[0061] Step B2: for each attribute set in the lexicographic order, verify whether B→f(g(B))-B holds in order from the smallest to the largest lexicographic order, where B is the attribute set in the lexicographic order; if B→f(g(B))-B holds, then continue to determine whether B is equal to f(g(B)); if B is equal to f(g(B)), then add B to the concept connotation set; if B is not equal to f(g(B)), then add B→f(g(B))-B to the value-dependent implication set;

[0062] Specifically, f() and g() are two operators in situational concept analysis. f() represents a common attribute operator, for example, if C represents any object set, f(C) represents the set of common attributes among all objects in the object set C; g() represents a common object operator, for example, if D represents any attribute set, g(D) represents the set of common objects corresponding to all attributes in the attribute set D.

[0063] f(g(B)) and B are two attribute sets, where f(g(B)) is the attribute set obtained by subjecting attribute set B to the g() operation and then to the f() operation; “f(g(B))-B” refers to the attribute set after removing the attributes in attribute set f(g(B)) that overlap with attribute set B; “B→f(g(B))-B” means that in the formal context, an object containing attribute set B must contain attribute set “f(g(B))-B”.

[0064] Step B3: After processing all attribute sets in the lexicographic order according to step B2, all value-dependent implication formulas in the value-dependent implication formula set are screened, and the screening condition is: deleting the value-dependent implication formulas whose value-dependent front key contains a decision attribute or whose value-dependent rear key contains a conditional attribute;

[0065] Step B4: For all the value-dependent implication formulas after screening, the attributes in the value-dependent implication formulas are matched with the crime names and / or criminal acts, that is, a series of value-dependent implication formulas corresponding to the criminal acts and the crime names are obtained.

[0066] For example, after processing the formal background shown in Table 1, the value-dependent implication formulas that can be obtained are a1→b1, a2a3→b2; among them, the implication formula a1→b1 means that if the criminal document contains the words such as stolen, pilfered, stolen, and stolen, the case is considered to be theft; the implication formula a2a3→b2 means that if the criminal document contains the words such as false invoice, purchase, purchased, and forgery, and contains invoices and invoice fees, the case is considered to be the crime of false invoicing.

[0067] S103: Process the case to be predicted based on natural language processing technology to obtain the criminal acts contained in the case to be predicted, match the criminal acts in the case to be predicted with a series of value-dependent implications, and obtain the crime corresponding to the criminal acts in the case to be predicted.

[0068] Specifically, the criminal behavior in the case to be predicted is matched with a series of the value-dependent implication formulas. If the criminal behavior successfully matches the first key in the value-dependent implication formula, the second key in the successfully matched value-dependent implication formula is the crime corresponding to the criminal behavior.

[0069] The crime prediction method based on situation concept analysis provided by the embodiment of the present invention is based on natural language processing and situation concept analysis (FCA) technology, and can quickly and accurately predict the crime of a case, thereby assisting relevant practitioners to quickly and accurately predict the crime based on the case.

[0070] Example 2

[0071] Corresponding to the above-mentioned crime prediction method based on situation concept analysis, such as Figure 2 As shown, the embodiment of the present invention also provides a crime prediction system based on situation concept analysis, including: a text processing module, a data analysis module and a crime prediction module; wherein:

[0072] The text processing module is used to process the existing case judgment documents and related laws and regulations based on natural language processing technology to obtain the criminal acts and charges in the existing case judgment documents and related laws and regulations. The data analysis module is used to process the criminal acts and charges in the existing case judgment documents and related laws and regulations into a formal background, and use a formal concept analysis method to process the formal background to generate a series of value-dependent implications corresponding to the criminal acts and charges. The charge prediction module is used to process the case to be predicted based on natural language processing technology to obtain the criminal acts contained in the case to be predicted, match the criminal acts in the case to be predicted with a series of value-dependent implications, and obtain the charges corresponding to the criminal acts in the case to be predicted.

[0073] As an implementable method, the data analysis module includes an attribute and object extraction unit and a formal background generation unit; wherein the attribute and object extraction unit is used to take the criminal behavior and crime as the attributes in the formal background, and take each existing case judgment document or relevant laws and regulations as an object in the formal background. The formal background generation unit is used to mark the intersection cell of the object and the attribute as 1 for each object when the object contains the corresponding attribute, otherwise it is marked as 0, until all objects are marked, and a table containing only 0, 1, object, and attribute is obtained, and the table is used as the formal background.

[0074] As an implementable method, the data analysis module further includes an attribute sorting unit, an attribute verification unit, a screening unit, and a value-dependent implication generation unit; wherein the attribute sorting unit is used to extract all attributes in the formal background and sort all attributes in the order of the lexicographic order; the attribute verification unit is used to verify whether B→f(g(B))-B holds true for each attribute set in the lexicographic order in the order from small to large, wherein B is the attribute set in the lexicographic order; if B→f(g(B))-B holds true, then continue to judge whether B→f(g(B))-B holds true. Is it equal to f(g(B)), if B is equal to f(g(B)), then add B to the concept connotation set; if B is not equal to f(g(B)), then add B→f(g(B))-B to the value-dependent implication set; the screening unit is used to screen all value-dependent implication formulas in the value-dependent implication set obtained by the attribute verification unit after processing all attribute sets in the lexicographic order, and the screening condition is: delete the value-dependent implication formulas whose value-dependent front key contains decision attributes or whose value-dependent rear key contains conditional attributes. The value-dependent implication formula generation unit is used to correspond the attributes in the value-dependent implication formulas with the crime and / or criminal behavior for all the value-dependent implication formulas after screening, and obtain a series of value-dependent implication formulas corresponding to the crime and the criminal behavior.

[0075] As an implementable method, the attribute sorting unit is specifically used for:

[0076] Define the order of magnitude between attributes, specifically: if all criminal acts are represented by a1, a2, a3...an, and all crimes are represented by b1, b2, b3...bn, then define the order of magnitude of all attributes as: a1<a2<a3<...<an<b1<b2<b3...<bn; then define the order of magnitude between attribute sets, specifically: set the set of all attributes G={a1,a2,a3...a n1 ,b1,b2,b3...b n2}, there is a property set and attribute collection The elements in the attribute sets A and B are sorted according to the definition of step B11. If the first unequal element in A and B starting from the left is larger than the first unequal element in B, then A is defined as < B. All attribute sets are sorted according to this definition, and the order from small to large is the lexicographical order. Each attribute set is different and satisfies the order from small to large in each attribute set.

[0077] The crime prediction method and system based on situation concept analysis provided by the present invention introduces the original text of relevant laws and regulations when processing case judgment documents, which can effectively improve the accuracy of crime prediction when facing uncommon cases.

[0078] Moreover, in judicial practice, criminal acts are often related, and the court's trial results often include multiple charges together. However, the prediction results of most current charge prediction methods can only obtain a single charge, and when faced with the prediction of multiple charges, the results are often not ideal. The value-dependent implication formula generated by the situation concept analysis in the present invention can be used for the prediction of multiple charges, which has great practical significance in assisting relevant practitioners.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A crime prediction method based on formal concept analysis, characterized in that: include: Step 1: Process the existing case judgment documents and relevant laws and regulations based on natural language processing technology to obtain the criminal acts and charges in the existing case judgment documents and relevant laws and regulations; Step 2: Treat the criminal acts and charges in the existing case judgment documents and relevant laws and regulations as formal background, and use the formal concept analysis method to process the formal background to generate a series of value-dependent implications containing criminal acts and charges; specifically, including: Step A1: Take the criminal act and crime as attributes in the formal background, and take each existing case judgment document or relevant laws and regulations as an object in the formal background; Step A2: For each object, if the object has the corresponding attribute, mark the intersection cell of the object and the attribute as 1, otherwise mark it as 0, until all objects are marked, and obtain a table containing only 0, 1, object, and attribute, which is the formal background; Step B1: extract all attributes in the formal background and sort all attributes in lexicographic order; Step B2: For each attribute set in the lexicographic order, verify whether B→f(g(B))-B holds in order from the smallest to the largest lexicographic order, where B is the attribute set in the lexicographic order; if B→f(g(B))-B holds, then continue to determine whether B is equal to f(g(B)); if B is equal to f(g(B)), then add B to the concept connotation set; if B is not equal to f(g(B)), then add B→f(g(B))-B to the value-dependent implication set; Step B3: After processing all attribute sets in the lexicographic order according to step B2, all value-dependent implication formulas in the value-dependent implication formula set are screened, and the screening condition is: deleting the value-dependent implication formulas whose value-dependent front key contains a decision attribute or whose value-dependent rear key contains a conditional attribute; Step B4: For all the value-dependent implication formulas that have been screened, the attributes in the value-dependent implication formulas are matched with the crime names and / or criminal acts, that is, a series of value-dependent implication formulas corresponding to the criminal acts and the crime names are obtained; Step 3: Process the case to be predicted based on natural language processing technology to obtain the criminal acts contained in the case to be predicted, match the criminal acts in the case to be predicted with a series of value-dependent implications, and obtain the crime corresponding to the criminal acts in the case to be predicted.

2. According to claim 1, a crime prediction method based on formal concept analysis is characterized in that: In step B1, all attributes are sorted in lexicographic order, specifically including: Step B11: Define the order of the attributes. Specifically, if all criminal acts are classified as a1, a2, a3...a n1 It means that all the crimes are classified into b1, b2, b3...b n2 Indicates that the order of the size of all attributes is defined as: a1<a2<a3<...<a n1 <b1<b2<b3<...<b n2 ; Step B12: define the size order between attribute sets, specifically: set the set of all attributes G = {a1, a2, a3...a n1, b1,b2,b3...b n2 }, there exist attribute sets A⊆G and attribute sets B⊆G, the elements in attribute sets A and B are sorted according to the definition in step B11, if the first unequal element in A and B from the left is larger and the first unequal element in B is smaller, then A<B; Step B13: sort all attribute sets according to the definition in step B12, and the obtained order from small to large is the lexicographic order; wherein each attribute set is different and satisfies the order from small to large in each attribute set.

3. A crime prediction system based on formal concept analysis, characterized in that: include: The text processing module is used to process the existing case judgment documents and relevant laws and regulations based on natural language processing technology to obtain the criminal acts and charges in the existing case judgment documents and relevant laws and regulations; The data analysis module is used to process the criminal acts and charges in the existing case judgment documents and relevant laws and regulations as formal backgrounds, and use formal concept analysis methods to process the formal backgrounds to generate a series of value-dependent implication formulas corresponding to the criminal acts and charges; The data analysis module includes an attribute and object extraction unit, a form background generation unit, an attribute sorting unit, an attribute verification unit, a screening unit and a value-dependent implication generation unit; The attribute and object extraction unit is used to take the criminal behavior and crime as attributes in the formal background, and take each existing case judgment document or relevant laws and regulations as an object in the formal background; The formal background generating unit is used for marking the intersection cell of each object and the attribute as 1 when the object has the corresponding attribute, otherwise it is marked as 0, until all objects are marked, and a table containing only 0, 1, object, and attribute is obtained, and the table is used as the formal background; an attribute sorting unit, for extracting all attributes in the formal background and sorting all attributes in lexicographic order; The attribute verification unit is used to verify whether B→f(g(B))-B holds for each attribute set in the lexicographic order in ascending order of the lexicographic order, where B is the attribute set in the lexicographic order; if B→f(g(B))-B holds, then continue to judge whether B is equal to f(g(B)); if B is equal to f(g(B)), then add B to the concept connotation set; if B is not equal to f(g(B)), then add B→f(g(B))-B to the value-dependent implication set; A screening unit is used to screen all value-dependent implication formulas in the value-dependent implication formula set obtained by the attribute verification unit after the attribute verification unit processes all attribute sets in the lexicographic order, wherein the screening condition is: deleting the value-dependent implication formulas whose value-dependent front key contains a decision attribute or whose value-dependent rear key contains a conditional attribute; A value-dependent implication generating unit is used to correspond the attributes in the value-dependent implication to the crime name and / or criminal behavior for all the value-dependent implication after screening, so as to obtain a series of value-dependent implication corresponding to the criminal behavior and the crime name; The crime prediction module is used to process the case to be predicted based on natural language processing technology, obtain the criminal behavior contained in the case to be predicted, match the criminal behavior in the case to be predicted with a series of value-dependent implications, and obtain the crime corresponding to the criminal behavior in the case to be predicted.

4. The crime prediction system based on formal concept analysis according to claim 3 is characterized in that: The attribute sorting unit is specifically used for: Define the order of magnitude between attributes. Specifically, if all criminal acts are assigned the order of a1, a2, a3...a n1 It means that all the crimes are classified into b1, b2, b3...b n2 Indicates that the order of the size of all attributes is defined as: a1<a2<a3<...<a n1 <b1<b2<b3<...<b n2 ; Define the size order between attribute sets, specifically: set the set of all attributes G = {a1, a2, a3...a n1, b1,b2,b3...b n2 }, there exist attribute sets A⊆G and attribute sets B⊆G, the elements in attribute sets A and B are sorted according to the definition in step B11, if the first unequal element in A and B from the left is larger and the first unequal element in B is smaller, then A<B; All attribute sets are sorted according to the definition, and the resulting order from small to large is the lexicographical order; wherein each attribute set is different and satisfies the order from small to large in each attribute set.

Images