Method for rapidly screening effective case information

Abstract

The invention provides a method for rapidly screening effective case information, and the method comprises the following steps: obtaining a bank evidence compression package, uploading the compression package to a library, screening out a to-be-decompressed object with the complete compression package, decompressing the screened to-be-decompressed object with the complete compression package, and storing the decompressed object with the complete compression package in a database. According to the method, a business context object is constructed to obtain information in a compressed package, whether the information is correct or not is checked, and the checking is that a random forest algorithm is utilized, non-corresponding information is marked, correct information is stored, and incorrect information is checked and modified manually. A trained random forest model is utilized to verify and correct the result of an identity card ocr, and the random forest model predicts by integrating a plurality of features (such as the length of an identity card number, the validity of a check code, the format of a name and the like) instead of depending on single ocr output, so that the overall accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of judicial industry data processing and storage technology, and in particular to a method for quickly screening effective case information. Background Art

[0002] "Legal collection" usually refers to debt collection through legal means, that is, when the debtor fails to repay the debt on time, the creditor uses legal means (such as litigation, arbitration, etc.) to recover the debt. In financial institutions, legal collection is a common debt collection method, especially when other collection methods are ineffective.

[0003] The current situation is that after the business obtains the compressed package of the original case information from the financial institution, the lawyer decompresses it himself and obtains n folders separated by case numbers, which include detailed billing tables, collection records, repayment records, application form pictures, ID card pictures, and photos of the applicant with bank personnel and other evidence. The lawyer then compares them with the naked eye to see whether the ID card is missing, whether the ID card is still valid, whether the application form is missing, whether the repayment record balance, interest, other fees, etc. are normal, and then judges whether the case is a valid case that can be used. This operation method that directly relies on manual inspection is relatively inefficient. One person can only handle about 20 cases a day. When handling a large number of financial cases, it takes a lot of staff time.

[0004] Therefore, it is necessary to provide a new method for quickly screening effective case information to solve the above technical problems. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a method for quickly screening effective case information.

[0006] The method for quickly screening effective case information provided by the present invention comprises the following steps:

[0007] Obtain the bank evidence compression package and upload it to the database. The bank evidence compression package includes ID card information, application form information, bill details and collection record information;

[0008] Filter out the objects to be decompressed that have complete compressed packages, and store the objects that have incomplete compressed packages directly in the database;

[0009] Decompress the complete objects to be decompressed in the filtered compressed package;

[0010] Construct a business context object to obtain the ID card information, application form information, bill details information and collection record information in the compressed package and verify whether the ID card information, application form information, bill details information and collection record information are correct. The verification uses the random forest algorithm to perform OCR recognition of the ID card information, OCR recognition of the application form information, parsing of the bill details information and parsing of the collection record information, and marks the ID card information, application form information, bill details information and collection record information that do not correspond;

[0011] The correct ID card information, application form information, billing details and collection record information are saved, and incorrect ID card information, application form information, billing details and collection record information are manually checked and modified.

[0012] Preferably, when the random forest algorithm recognizes the ID card information OCR, the following steps are included:

[0013] S1. Data preparation: collect ID card information;

[0014] S2. Build and train a random forest model;

[0015] S3. Use the trained random forest model to verify and correct the ID card OCR results.

[0016] Preferably, when the random forest algorithm is used for OCR recognition of ID card information, training the random forest model includes the following steps:

[0017] S401, collecting training data: collecting known correct ID card information as training samples;

[0018] S402, feature extraction: extracting field features from OCR output;

[0019] S403, constructing feature vectors: organizing the extracted features into feature vectors, each feature vector corresponds to a sample, and marking each field;

[0020] S404, classification: classify labels according to field tags and indicate whether they are correct;

[0021] S405, training a random forest model: using the feature vectors and corresponding labels to train a random forest model.

[0022] Preferably, the ID card OCR recognition includes five dimensions of information, namely name, gender, nationality, date of birth and ID card number. When any dimension of information is incorrect, the dimension is marked.

[0023] Preferably, the application form OCR recognition includes six dimensions of information, namely, applicant signature, application date, contact information, loan amount, repayment period and barcode number. When any dimension of information is incorrect, the dimension is marked.

[0024] Preferably, the bill detail analysis includes five dimensions of information, namely, balance, interest, fees, transaction date, and transaction amount. When any dimension of information is incorrect, the dimension is marked.

[0025] Preferably, the collection record analysis includes four dimensions of information, namely, collection date, collection amount, collection description and collection times, wherein when any dimension of information is incorrect, the dimension is marked.

[0026] Compared with the related art, the method for quickly screening effective case information provided by the present invention has the following beneficial effects:

[0027] 1. When screening case information, the present invention performs conditional judgment on the compressed package to screen out the complete objects to be decompressed in the compressed package, and then obtains the case information by constructing a business context object and verifies whether the case information is correct. Verifying the case information includes using a random forest algorithm to perform OCR recognition on identity cards, OCR recognition on application forms, parsing of bill details, and parsing of collection records, and marking incorrect case information. When manually checking case information, it is only necessary to check the marked information without comparing one by one, thereby greatly reducing the workload of the staff and greatly increasing the number of cases manually checked per unit time. To a certain extent, the screening efficiency of case information is improved.

[0028] 2. The present invention uses a random forest algorithm, which can process a large number of input features and automatically perform feature selection. The present invention uses a trained random forest model to verify and correct the results of the ID card OCR. The random forest model makes predictions by integrating multiple features (such as the length of the ID card number, the validity of the check code, the format of the name, etc.), rather than relying solely on a single OCR output, thereby helping to improve the overall accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A schematic diagram of a process for quickly screening effective case information provided by the present invention;

[0030] Figure 2 A schematic flow chart of the steps of the random forest algorithm provided by the present invention;

[0031] Figure 3A flowchart of the steps of training a random forest model provided by the present invention. DETAILED DESCRIPTION

[0032] The present invention will be further described below in conjunction with the accompanying drawings and implementation modes.

[0033] Please refer to Figure 1 — Figure 3 ,in, Figure 1 A schematic diagram of the structure of the method for quickly screening effective case information provided by the present invention; Figure 2 A schematic diagram of the structure of the steps of the random forest algorithm provided by the present invention; Figure 3 A structural schematic diagram of the steps of training a random forest model provided by the present invention.

[0034] In the specific implementation process, a method for quickly screening effective case information is structured as follows: Figure 1 — Figure 3 As shown, the following steps are included:

[0035] Obtain the bank evidence compression package and upload it to the database. The bank evidence compression package includes ID card information, application form information, bill details and collection record information;

[0036] Filter out the objects to be decompressed that have complete compressed packages, and store the objects that have incomplete compressed packages directly in the database;

[0037] Decompress the complete objects to be decompressed in the filtered compressed package;

[0038] Construct a business context object to obtain the ID card information, application form information, bill details information and collection record information in the compressed package and verify whether the ID card information, application form information, bill details information and collection record information are correct. The verification uses the random forest algorithm to perform OCR recognition of the ID card information, OCR recognition of the application form information, parsing of the bill details information and parsing of the collection record information, and marks the ID card information, application form information, bill details information and collection record information that do not correspond;

[0039] Save the correct ID card information, application form information, bill details and collection record information, and manually check and modify the incorrect ID card information, application form information, bill details and collection record information;

[0040] The random forest algorithm includes the following steps when performing OCR recognition on ID card information:

[0041] S1. Data preparation: collect ID card information;

[0042] S2. Build and train a random forest model;

[0043] S3, use the trained random forest model to verify and correct the ID card OCR results;

[0044] When the random forest algorithm is used for OCR recognition of ID card information, training the random forest model includes the following steps:

[0045] S401, collecting training data: collecting known correct ID card information as training samples;

[0046] S402, feature extraction: extract field features from the OCR output, such as name, gender, ethnicity, date of birth, ID number, etc., where:

[0047] When checking a name, check the following:

[0048] ① Character length: the number of characters in the name;

[0049] ② Whether the first character is a Chinese character: Check whether the first character of the name is a Chinese character;

[0050] ③Whether there are special characters: Check whether the name contains special characters, English characters and numbers.

[0051] When checking gender, you can check the following:

[0052] ① Is it "Male" or "Female": Check whether the gender field is one of these two values;

[0053] ②Character length: the number of characters for gender (should be 1 or 2)

[0054] ③Whether it is Chinese characters: Check whether the gender field is Chinese characters;

[0055] When checking ethnicity, check the following:

[0056] ① Character length: the number of characters of the nation;

[0057] ② Is it a common ethnic group: Check whether the ethnic group field is one of the common ethnic groups in China (such as Han, Mongolian, etc.);

[0058] ③Whether it is Chinese characters: Check whether the ethnicity field is Chinese characters.

[0059] When checking your date of birth, you can check the following:

[0060] ①Whether the format is correct: Check whether the date of birth conforms to the format of YYYY-MM-DD or YYYYMMDD;

[0061] ②Year range: Check whether the year is within a reasonable range (e.g. 1900 to the current year);

[0062] ③Month range: check whether the month is between 1 and 12;

[0063] ④Date range: Checks whether the date is between 1 and 31 and does not exceed the maximum number of days in a specific month (such as February);

[0064] ⑤ Whether there are illegal characters: Check whether the date of birth contains illegal characters (such as letters, special symbols, etc.);

[0065] When checking your ID number, check the following:

[0066] ① Is the length 18 digits: Check whether the ID card number is exactly 18 digits;

[0067] ② Whether the first 6 digits correspond to the legal administrative division code: Check whether the first 6 digits correspond to the real administrative division code of China;

[0068] ③Whether the 7th to 14th digits are the date of birth: Check whether the 7th to 14th digits are consistent with the date of birth field;

[0069] ④Whether the 15th to 17th digits are sequence codes: Check whether the 15th to 17th digits are reasonable sequence codes (odd numbers are assigned to males, and even numbers are assigned to females);

[0070] ⑤ Is the check code correct? Calculate the check code based on the first 17 digits and check whether the 18th digit is consistent with the calculated result;

[0071] ⑥Whether there are illegal characters: Check whether the ID number contains illegal characters (such as letters, special symbols, etc.)

[0072] S403, constructing feature vectors: organizing the extracted features into feature vectors, each feature vector corresponds to a sample, and marking each field;

[0073] S404, classification: classify labels according to field tags and indicate whether they are correct;

[0074] S405, training a random forest model: using the feature vectors and corresponding labels to train a random forest model;

[0075] The ID card OCR recognition includes five dimensions of information, namely name, gender, nationality, date of birth and ID card number. When any one of the dimensions is incorrect, the dimension is marked. The application form OCR recognition includes six dimensions of information, namely applicant signature, application date, contact information, loan amount, repayment period and barcode number. When any one of the dimensions is incorrect, the dimension is marked. The bill detail analysis includes five dimensions of information, namely balance, interest, fees, transaction date and transaction amount. When any one of the dimensions is incorrect, the dimension is marked. The collection record analysis includes four dimensions of information, namely collection date, collection amount, collection description and collection times. When any one of the dimensions is incorrect, the dimension is marked.

[0076] The working principle provided by the present invention is as follows: when screening case information, the present invention performs conditional judgment on the compressed package to screen out the complete objects to be decompressed in the compressed package, and then obtains the case information by constructing a business context object and verifies whether the case information is correct, wherein verifying the case information includes using a random forest algorithm to perform OCR recognition on identity cards, OCR recognition on application forms, parsing of bill details, and parsing of collection records, and marking incorrect case information. When manually checking case information, it is only necessary to check the marked information without comparing one by one, thereby greatly reducing the workload of staff and greatly increasing the number of cases manually checked per unit time, thereby improving the screening efficiency of case information to a certain extent.

[0077] The circuits and controls involved in the present invention are all prior art and will not be described in detail here.

[0078] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for quickly screening effective case information, characterized in that: The following steps are involved: Obtain the bank evidence compression package and upload it to the database. The bank evidence compression package includes ID card information, application form information, bill details and collection record information; Filter out the objects to be decompressed that have complete compressed packages, and store the objects that have incomplete compressed packages directly in the database; Decompress the complete objects to be decompressed in the filtered compressed package; Construct a business context object to obtain the ID card information, application form information, bill details information and collection record information in the compressed package and verify whether the ID card information, application form information, bill details information and collection record information are correct. The verification uses the random forest algorithm to perform OCR recognition of the ID card information, OCR recognition of the application form information, parsing of the bill details information and parsing of the collection record information, and marks the ID card information, application form information, bill details information and collection record information that do not correspond; The correct ID card information, application form information, billing details and collection record information are saved, and incorrect ID card information, application form information, billing details and collection record information are manually checked and modified.

2. The method for quickly screening effective case information according to claim 2, characterized in that: The random forest algorithm includes the following steps when performing OCR recognition on ID card information: S1. Data preparation: collect ID card information; S2. Build and train a random forest model; S3. Use the trained random forest model to verify and correct the ID card OCR results.

3. The method for quickly screening effective case information according to claim 3, characterized in that: When the random forest algorithm is used for OCR recognition of ID card information, training the random forest model includes the following steps: S401, collecting training data: collecting known correct ID card information as training samples; S402, feature extraction: extracting field features from OCR output; S403, constructing feature vectors: organizing the extracted features into feature vectors, each feature vector corresponds to a sample, and marking each field; S404, classification: classify labels according to field tags and indicate whether they are correct; S405, training a random forest model: using the feature vectors and corresponding labels to train a random forest model.

4. The method for quickly screening effective case information according to claim 1, characterized in that: The ID card OCR recognition includes five dimensions of information, namely name, gender, nationality, date of birth and ID card number. When any dimension of information is incorrect, the dimension will be marked.

5. The method for quickly screening effective case information according to claim 1, characterized in that: The application form OCR recognition includes six dimensions of information, namely, applicant signature, application date, contact information, loan amount, repayment period and barcode number. If any dimension of information is incorrect, the dimension will be marked.

6. The method for quickly screening effective case information according to claim 1, characterized in that: The bill detail analysis includes five dimensions of information, namely balance, interest, fees, transaction date and transaction amount. When any dimension information is incorrect, the dimension will be marked.

7. The method for quickly screening effective case information according to claim 1, characterized in that: The collection record analysis includes four dimensions of information, namely, collection date, collection amount, collection description and collection times. When any dimension of information is incorrect, the dimension is marked.

Images