A government affair service matter recommendation method and system

Abstract

The application discloses a government affair service matter recommendation method and system, and relates to the field of data processing, and solves the problems of low recommendation accuracy and high implementation cost in the prior art. The government affair service matter recommendation method comprises the following steps: obtaining a basic database and the identity of a user, and then obtaining a plurality of historical similar case records; performing seasonal effect weighting on the plurality of historical similar case records respectively, and obtaining a plurality of corresponding weighted historical similar case records; if the user has a historical case record, performing sequence pattern mining on the plurality of weighted historical similar case records, and then obtaining a recommendation list as recommendation information output; otherwise, based on the plurality of weighted historical similar case records, obtaining a plurality of corresponding handling frequencies and a plurality of corresponding confidence degrees, and then obtaining a recommendation list as recommendation information output. The application is suitable for the field of e-government intelligence.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically to the field of personalized recommendation. Background Technology

[0002] With the deepening of "Internet + Government Services," government service platforms have accumulated massive amounts of user case data. To improve service efficiency, personalized recommendation technology has gradually been introduced into the government service sector. Existing government service recommendation technologies mainly include deep learning-based recommendation methods, time-matching-based recommendation methods, and multimodal data-based recommendation methods, as detailed below: First, regarding recommendation methods based on deep learning, Chinese patent document CN107368549A discloses a "Personalized Government Service Recommendation Method and System Based on Deep Learning." This technical solution establishes a personalized feature pattern library by collecting user behavior data and uses a deep learning model to match user needs with government services. However, this type of method relies on a large amount of historical behavior data for model training, and suffers from a cold start problem for new users or newly introduced policies.

[0003] Second, regarding time-matching recommendation methods, Chinese patent document CN117216403A discloses "A Web-based Personalized Service Recommendation Method." This technical solution discloses a method for recommending government resources based on pre-aggregation and optimization, calculating the matching degree according to attributes such as the release time and expiration time of government resources. This type of method focuses on resource timeliness management but does not consider the temporal correlation of users' historical behavior sequences, resulting in low recommendation accuracy.

[0004] Third, regarding recommendation methods based on multimodal data, Chinese patent document CN121352938A discloses a "Multimodal Member Identity Recognition and Service Recommendation System Based on a Large Model." This technical solution uses multimodal information such as visual data, behavioral data, and text data for identity recognition and recommendation. Such methods require complex multimodal data acquisition equipment, resulting in high implementation costs, and they are not optimized for the unique characteristics of government services, such as cross-domain dynamic changes in service standards and concentrated seasonal demand surges.

[0005] In summary, existing government service recommendations suffer from problems such as cold start, failure to consider users' historical behavior sequences, low recommendation accuracy, and high implementation costs. Summary of the Invention

[0006] This invention alleviates the problems of cold start, failure to consider users' historical behavior sequences, low recommendation accuracy, and high implementation cost in existing government service recommendations.

[0007] This invention provides the following solution: Option 1: A method for recommending government service items, comprising the following steps: Data collection steps include obtaining the basic database and user identities; The data processing steps are as follows: based on the identity, obtain the corresponding identity features; based on the identity features, obtain augmented data conditions; based on the augmented data conditions and the basic database, obtain several historical similar case records. The seasonal effect weighting step involves applying seasonal effect weighting to the aforementioned historical similar case records to obtain corresponding weighted historical similar case records. In the recommendation strategy selection step, if the user has a history of handling cases, then the historical user recommendation list generation step is performed; otherwise, the new user recommendation list generation step is performed. The steps for generating a historical user recommendation list include: performing sequence pattern mining on the several weighted historical similar case records to obtain frequent sequences; and generating a recommendation list as recommendation information output based on the historical case records and the frequent sequences. The steps for generating a new user recommendation list are as follows: based on the weighted historical similar case records, obtain the corresponding number of processing frequencies and number of confidence levels; sort the weighted historical similar case records based on the number of processing frequencies and the number of confidence levels to obtain a recommendation list as the output recommendation information.

[0008] Furthermore, in one embodiment of the present invention, in the data acquisition step, the basic database includes an event database; The aforementioned database is used to store business processing item codes and reused business codes; The service item code is used to identify items within the same region; The reused service code is used to identify the same item across regions.

[0009] Furthermore, in one embodiment of the present invention, in the data processing step, the number of the plurality of historical similar case records is greater than the data volume threshold.

[0010] Furthermore, in one embodiment of the present invention, the method for weighting the seasonal effect in the seasonal effect weighting step is as follows: Based on the current month and historical frequency of occurrence, the support score of the historical similar case records is obtained, and the support score and the corresponding historical similar case records are used as weighted historical similar case records. The historical occurrence frequency refers to the historical occurrence frequency of the historical similar case records.

[0011] Furthermore, in one embodiment of the present invention, the support is obtained by weighting the current month and the historical frequency of occurrence.

[0012] Furthermore, in one embodiment of the present invention, the method for sequence pattern mining in the historical user recommendation list generation step is as follows: Based on the aforementioned event database, a global event sequence is obtained; Using the items from the user's historical completed cases in the basic database as a prefix, a projection database is constructed based on the global item sequence; The frequent sequence patterns are recursively mined in the projection database until the minimum support threshold is met. All frequent sequence patterns that meet the minimum support threshold are considered frequent sequences.

[0013] Furthermore, in one embodiment of the present invention, the method further includes a cold start step for newly established or newly introduced matters. If the newly introduced matters match the aforementioned identity characteristics. Or, the semantic similarity between the newly introduced item and the recommended list exceeds a semantic threshold, and the recommended list is a list of recommendations output as recommendation information; Or the newly introduced matters comply with policy promotion rules. The newly introduced items will then be added to the recommended list.

[0014] Option 2: A government service recommendation system, the system comprising the following modules: The data acquisition module obtains the basic database and user identity. The data processing module obtains corresponding identity features based on the identity; obtains augmented data conditions based on the identity features; and obtains several historical similar case records based on the augmented data conditions and the basic database. The seasonal effect weighting module applies seasonal effect weighting to the aforementioned historical similar case records to obtain corresponding weighted historical similar case records. The recommendation strategy selection module, if the user has a history of processed cases, then performs the historical user recommendation list generation module; otherwise, it performs the new user recommendation list generation module. The historical user recommendation list generation module performs sequence pattern mining on the several weighted historical similar case records to obtain frequent sequences; based on the historical case records and the frequent sequences, it obtains a recommendation list as recommendation information output. The new user recommendation list generation module obtains several processing frequencies and several confidence levels based on several weighted historical similar case records; it sorts the several weighted historical similar case records based on the several processing frequencies and several confidence levels, and outputs a recommendation list as recommendation information.

[0015] Option 3: An electronic device, comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus; Memory, used to store computer programs; When a processor executes a computer program stored in memory, it implements the method described in Scheme 1.

[0016] Option 4: A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the method described in Option 1.

[0017] The government service item recommendation method and system described in this invention effectively alleviate the problems of cold start, lack of consideration for users' historical behavior sequences, low recommendation accuracy, and high implementation costs in existing government service item recommendation systems. Specific beneficial effects include: This invention describes a method for recommending government service items. It establishes a unified identification mechanism for cross-regional items, identifying the same item across different regions and time periods, thus avoiding scattered and duplicated recommendations, and preventing disruptions to the continuity of user behavior sequences due to changes in item names. A dynamic weight adjustment mechanism is designed to address the seasonal effect of government services, prioritizing seasonal government services for recommendation at appropriate times. However, it lacks a multi-level cold start mechanism encompassing identity classification, semantic similarity, and rule-driven approaches, providing accurate recommendations for cold start scenarios such as newly registered users with no historical processing records or newly introduced items without historical processing data.

[0018] The method described in this invention is applicable to the field of intelligent e-government. Attached Figure Description

[0019] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of the government service item recommendation method described in Implementation Method 1; Figure 2 This is the recommended flowchart described in Implementation Method 8. Detailed Implementation

[0020] Various embodiments of the present invention will now be clearly and completely described with reference to the accompanying drawings. The embodiments described with reference to the drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0021] Implementation Method 1: A method for recommending government service items as described in this implementation method, such as... Figure 1 As shown, it includes the following steps: Data collection steps include obtaining the basic database and user identities; The data processing steps are as follows: based on the identity, obtain the corresponding identity features; based on the identity features, obtain augmented data conditions; based on the augmented data conditions and the basic database, obtain several historical similar case records. The seasonal effect weighting step involves applying seasonal effect weighting to the aforementioned historical similar case records to obtain corresponding weighted historical similar case records. In the recommendation strategy selection step, if the user has a history of handling cases, then proceed to the step of generating a historical user recommendation list; otherwise, proceed to the step of generating a new user recommendation list. The steps for generating a historical user recommendation list include: performing sequence pattern mining on the several weighted historical similar case records to obtain frequent sequences; and generating a recommendation list as recommendation information output based on the historical case records and the frequent sequences. The steps for generating a new user recommendation list are as follows: based on the weighted historical similar case records, obtain the corresponding number of processing frequencies and number of confidence levels; sort the weighted historical similar case records based on the number of processing frequencies and the number of confidence levels to obtain a recommendation list as the output recommendation information.

[0022] In this embodiment, the sorting of the weighted historical similar case records based on the processing frequency and the confidence level is specifically as follows: first, sorting them in descending order according to the processing frequency to obtain a recommendation list (the first few items are not taken from the recommendation list, and the complete list is maintained); then, sorting the items in the recommendation list according to the confidence level, and taking the top N items from the confidence level sorting results as the recommendation results.

[0023] In this embodiment, the recommendation method for historical users is based on a sequence pattern mining algorithm. First, frequent sequence patterns are obtained, and then subsequent items are obtained by matching these frequent sequence patterns with the user's historical application records. The preferred sequence pattern mining algorithm is the PrefixSpan algorithm.

[0024] In this embodiment, the user's identity in the data collection step includes both natural persons and legal persons. By identifying the current user as a natural person or legal person, the corresponding identity feature data is extracted; based on the extracted identity feature data, the historical case processing database is queried to obtain the user's historical case processing records.

[0025] In this embodiment, during the data processing step, a similar user group is obtained based on the augmented data conditions and the basic database; based on the augmented data conditions, several historical similar case records of the similar user group are obtained from the basic database.

[0026] In this embodiment, the augmented data conditions in the data processing step include: When the user's category is natural person, extract age, gender, ethnicity, and personal identity characteristics of urban and rural residents' pension insurance contributions. Expand the age characteristic forward and backward by a preset number of years, expand the ethnicity characteristic to include similar ethnic categories, and expand the personal identity characteristics of urban and rural residents' pension insurance contributions to include the same identity category. When the user's category is a legal person, extract the characteristics of industry category, business period and total tax amount, expand the business period to the front and back of a preset number of years, and adjust the total tax amount up and down according to a preset ratio. Based on the expanded combination of identity features, query the historical processing records of similar user groups; When the amount of data in the query results is lower than the minimum data threshold required for sequence pattern mining, the identity feature conditions are gradually reduced according to the preset priority until the amount of data in the query results meets the conditions for sequence pattern mining.

[0027] In this embodiment, in the step of generating the historical user recommendation list, the user's historical case records are used to search for subsequent items in the frequent sequence to generate the recommendation list.

[0028] In this embodiment, the recommendation list is sorted in descending order of frequency, and the top 3 items are output to the user.

[0029] Implementation Method 2: This implementation method further defines the government service item recommendation method described in Implementation Method 1. In this implementation method, the basic database in the data collection step includes an item database. The aforementioned database is used to store business processing item codes and reused business codes; The service item code is used to identify items within the same region; The reused service code is used to identify the same item across regions.

[0030] In this embodiment, the basic database also includes a population database, a legal entity database, and a database of historical case acceptance.

[0031] The population database stores natural person identity data; the legal person database stores legal person identity data; and the case handling history database stores user's historical case handling records.

[0032] The population database includes the following fields: ID number, gender, age, ethnicity, and individual identity as a contributor to the urban and rural residents' basic pension insurance scheme. The individual identity as a contributor to the urban and rural residents' basic pension insurance scheme includes workers, cadres, farmers, and clerks.

[0033] The legal entity database includes the following fields: Unified Enterprise Credit Code, Industry Category Name, Operating Period, and Total Tax Payment.

[0034] The historical application database includes the following fields: applicant's ID number, application time, application name, applicant's name, and application code.

[0035] The database of items includes the following fields: business processing item code, reused business code, item name, and item former name.

[0036] The business processing item code is used to uniquely identify items within the same region, and the reused business code is used to uniquely identify the same item across regions; the same item corresponds to different business processing item codes in different regions, but corresponds to the same reused business code; when the item name changes, the business processing item code and the reused business code remain unchanged.

[0037] For example, if it is a birth registration matter, as shown in Table 1, the business processing item codes are different in different regions, but the reused business codes are the same. Moreover, when the name of the matter changes from "first and second child birth registration" to "first, second, and third child birth registration", the reused business codes remain unchanged to ensure the continuity of historical records.

[0038] Table 1

[0039] This implementation further defines the data collection steps and explains the basic database. By reusing business codes, the same item from different regions, periods, and with different names is uniformly identified, constructing a global item sequence. Even if the item name changes due to policy adjustments, historical records can still be associated based on the unchanging reusable business codes. This solves the problem of user behavior sequence breaks caused by item name changes, enables cross-regional and cross-time period user behavior pattern mining, unifies cross-regional items, and ensures the continuity of behavior sequences.

[0040] Implementation Method 3: This implementation method further defines the government service item recommendation method described in Implementation Method 1. In this implementation method, during the data processing step, the number of the plurality of historical similar case records is greater than the data volume threshold.

[0041] In this embodiment, the data volume threshold is preferably 500.

[0042] Implementation Method Four: This implementation method further defines the government service recommendation method described in Implementation Method One. In this implementation method, the seasonal effect weighting step uses the following method: Based on the current month and historical frequency of occurrence, the support score of the historical similar case records is obtained, and the support score and the corresponding historical similar case records are used as weighted historical similar case records. The historical occurrence frequency refers to the historical occurrence frequency of the historical similar case records.

[0043] In this embodiment, the seasonal effect weights of each item in different months are preset; according to the current month, the corresponding seasonal effect weights are assigned to each item in the historical case records obtained from the query.

[0044] In this embodiment, the seasonal effect weight is dynamically calculated based on the historical monthly processing distribution of the event; for seasonal events with obvious periodicity, a weight higher than the historical average frequency is assigned to the current month.

[0045] In this implementation method, taking "college entrance examination registration" as an example, the historical monthly processing distribution is as follows: 10 times each from January to May, 200 times in June (college entrance examination season), and 5 times each from July to December. The total number of processing times is 300.

[0046] Dynamic calculation: The weight for June is 200÷(300÷12)=8.0, and the weight for other months is 10÷(300÷12)=0.4 or 5÷(300÷12)=0.2.

[0047] The current month is June. If there are any "college entrance examination registration" items in the query history, they will be assigned a weight of 8.0, which will significantly improve their ranking in the recommendation list.

[0048] This implementation further defines the seasonal effect weighting step and explains the seasonal effect weighting method. This method pre-determines the seasonal effect weights for each item in different months and dynamically weights historical case records. For seasonal items, a higher weight is assigned to the current corresponding month, ensuring that seasonal government affairs are prioritized for recommendation at appropriate times, effectively improving the timeliness of the recommendation results.

[0049] Implementation Method 5: This implementation method further defines the government service item recommendation method described in Implementation Method 4. In this implementation method, the support is obtained by weighting the current month and the historical frequency of occurrence.

[0050] Implementation Method Six: This implementation method further defines the government service item recommendation method described in Implementation Method Two. In this implementation method, the sequence pattern mining method in the historical user recommendation list generation step is as follows: Based on the aforementioned event database, a global event sequence is obtained; Using the items from the user's historical completed cases in the basic database as a prefix, a projection database is constructed based on the global item sequence; The frequent sequence patterns are recursively mined in the projection database until the minimum support threshold is met. All frequent sequence patterns that meet the minimum support threshold are considered frequent sequences.

[0051] In this embodiment, the method for obtaining the global event sequence is as follows: based on the reuse business code in the event database, the same events across regions and different periods are uniformly identified to construct a global event sequence; for historical records where the event name changes but the reuse business code remains unchanged, they are associated with the same event based on the reuse business code to ensure the continuity of the user behavior sequence.

[0052] In this embodiment, the method for obtaining the projection database is as follows: since the same item may have different names under different conditions in different regions and at different times, a unified name is obtained by using a reused business code. The projection database is obtained by replacing the original item name with the reused business code.

[0053] In this implementation, based on the reused business codes in the item database, the same items across regions and different periods are uniformly identified to construct a global item sequence; for historical records where the item name changes but the reused business code remains the same, they are associated with the same item based on the reused business code to ensure the continuity of the user behavior sequence.

[0054] Implementation Method Seven: This implementation method further defines the government service item recommendation method described in Implementation Method One. In this implementation method, the method also includes a cold start step for newly established or newly introduced items. If the newly introduced matters match the aforementioned identity characteristics. Or, if the semantic similarity between the newly introduced item and the recommended list exceeds a semantic threshold, the recommended list is a recommended list output as recommendation information; Or the newly introduced matters comply with policy promotion rules. The newly introduced items will then be added to the recommended list.

[0055] In this embodiment, it is determined whether newly issued items need to be added to the recommendation list; if so, the newly issued items are added to the recommendation list based on the cold start strategy for items; the cold start strategy for items includes identity classification matching, semantic similarity calculation and policy rule driving.

[0056] The determination of whether new items need to be added to the recommendation list includes identity classification matching, semantic similarity matching, and policy-driven matching; The identity classification and matching process involves manually labeling newly issued items with applicable user identity characteristics. When the current user identity characteristics match the labeled applicable identity characteristics, the newly issued item is added to the recommendation list. The semantic similarity matching involves performing semantic parsing and word embedding on newly issued items, calculating the semantic similarity with existing items in the recommendation list, and adding the newly issued item to the recommendation list when the semantic similarity exceeds a preset threshold. The policy-driven matching refers to adding newly introduced items that meet the pre-defined policy promotion rules, combined with user identity characteristics or historical case handling information, to the recommendation list.

[0057] This implementation further refines the method for recommending government services, explaining the cold start steps. This method addresses the challenges of recommending new users and new services. For newly introduced services, it constructs a three-tiered cold start strategy encompassing identity classification matching, semantic similarity matching, and rule-driven matching. For newly registered users, it performs frequency-based recommendations based on seasonally weighted data of similar user groups. The synergistic effect of these two-tiered cold start mechanisms effectively solves the cold start problem in government service recommendations.

[0058] Implementation Method 8: This implementation method is the government service item recommendation method described in Implementation Method 1, combined with the optimized steps of Implementation Methods 2 to 7.

[0059] In this embodiment, such as Figure 2 As shown, the recommendation process for individual user "Ms. Zhang" (26 years old, Han nationality, employee) is as follows: Step S1, Identity Recognition and Data Query: The system identifies Ms. Zhang as a natural person and extracts her identity characteristics: age 26, female, Han nationality, and employee with pension insurance status. The system queries the historical application database to retrieve her historical application records: ["Social Security Card Activation", "Social Security Payment Inquiry"].

[0060] Step S2, data condition augmentation and seasonal effect weighting: Data criteria were broadened: the age range was expanded to 21-31 years old (±5 years), the ethnicity remained Han, and the pension insurance status was expanded to "employee / cadre". Based on the expanded criteria, historical application records of similar user groups (Han female employees / cadres aged 21-31) were retrieved, yielding 1000 records.

[0061] If the query results are insufficient (e.g., less than 500 results), gradually reduce the conditions: first broaden the pension insurance status to all types, then broaden the age range to 16-45 years old, until the data volume meets the conditions for sequence pattern mining.

[0062] Seasonal weighting: The current month is March. The weight of "social security payment" is assumed to be 1.5 for January to March and 1.0 for other months. The "social security payment" records in the query results are assigned a weight of 1.5, and the weighted support is calculated.

[0063] Step S3: Determine user type and select recommendation strategy: Ms. Zhang has a history of handling cases, proceed to step S4.

[0064] Step S4, Sequence Pattern Mining and Item Recommendation: S401 first constructs a global event sequence before sequence pattern mining: based on reused business codes, the same events in different regions and at different times are uniformly identified. For example, "social security card activation" has different business processing codes in different regions, but the reused business code is "TE22000000000000222072301800002", which is uniformly identified as the same event, thus constructing a global event sequence.

[0065] S402 Sequence Pattern Mining and Item Recommendation: Using Ms. Zhang's historical records ["Social Security Card Activation", "Social Security Payment Inquiry"] as prefixes, a projected database is constructed based on global item sequences. The PrefixSpan algorithm is used to recursively mine frequent sequence patterns, with a minimum support threshold set to 0.01. Frequent patterns discovered: ["Social Security Card Activation", "Social Security Payment Inquiry"] → ["Social Security Card Replacement"] (Support 66, Confidence 0.30).

[0066] Based on this frequent pattern, a recommendation list is generated: ["Social Security Card Replacement", "Medical Insurance Payment", "Housing Provident Fund Inquiry"].

[0067] Step S5, Cold Start Processing: Determine if a new item needs to be added. Assume a new item, "Electronic Social Security Card Application," is introduced: (1) Identity Classification Matching: This item is marked as applicable to the identity of "employee / cadre". Ms. Zhang's identity matches and is added to the recommendation list; (2) Semantic similarity matching: The semantic similarity between "electronic social security card application" and "social security card replacement" is calculated to be 0.85, which exceeds the threshold of 0.8, so it is added to the recommendation list; (3) Policy-driven matching: This item is a government-promoted item and is added to the recommendation list.

[0068] The final recommended list is: "Social Security Card Replacement", "Electronic Social Security Card Application", "Medical Insurance Payment", and "Housing Provident Fund Inquiry".

[0069] Step S6, Recommendation Result Sorting and Output: Sorted by confidence level, the sorted results are: Electronic Social Security Card Application (0.85), Social Security Card Replacement (0.30), Medical Insurance Payment (0.25), Housing Provident Fund Inquiry (0.20). The first 3 items are output to the user terminal.

[0070] In this implementation, the recommended process for a newly registered corporate user, "a catering company" (with no prior application records), is as follows: Step A1, Identity Verification and Data Query: The system identifies the user as a legal entity and extracts the following identity features: industry category "accommodation and catering", business period start date "2022-08-15", and total tax payment of 13,200 yuan. A query of the historical application database reveals no historical application records.

[0071] Step A2, data condition augmentation and seasonality weighting: Data criteria were broadened: the operating period was extended to 2017-2027 (±5 years), and the total tax payment was extended to RMB 10,560-15,840 (±20%). Based on the expanded criteria, historical application records of similar user groups (accommodation and catering industry, registered between 2017 and 2027, tax payment of RMB 10,560-15,840) were retrieved, yielding 800 records.

[0072] Seasonal effect weighting: The current month is May. The "Renewal of Food Business License" item is assumed to have a weight of 1.3 in May and June (peak renewal period), and 1.0 in other months.

[0073] Step A3: Determine user type and select recommendation strategy: This enterprise has no historical application records, proceed to step S7.

[0074] Step A4, User Cold Start Recommendation: For the historical transaction records of similar user groups (accommodation and catering industry) obtained in Step S3 after seasonal effect weighting, the transaction frequency of each item is calculated, as shown in Table 2: Table 2

[0075] Step A5, Recommendation result sorting and output: Sort by frequency in descending order, and output the top 3 items to the user terminal: "Renewal of Food Business License", "Tax Declaration", "Social Security Account Opening".

Claims

1. A method for recommending government service items, characterized in that, Includes the following steps: Data collection steps include obtaining the basic database and user identities; The data processing steps are as follows: based on the identity, obtain the corresponding identity features; based on the identity features, obtain augmented data conditions; based on the augmented data conditions and the basic database, obtain several historical similar case records. The seasonal effect weighting step involves applying seasonal effect weighting to the aforementioned historical similar case records to obtain corresponding weighted historical similar case records. In the recommendation strategy selection step, if the user has a history of handling cases, then proceed to the step of generating a historical user recommendation list; otherwise, proceed to the step of generating a new user recommendation list. The step of generating a historical user recommendation list involves performing sequence pattern mining on the several weighted historical similar case records to obtain frequent sequences. Based on the historical case records and the frequent sequences, a recommendation list is obtained as the recommendation information output. The steps for generating a new user recommendation list are as follows: based on the weighted historical similar case records, obtain the corresponding number of processing frequencies and number of confidence levels; sort the weighted historical similar case records based on the number of processing frequencies and the number of confidence levels to obtain a recommendation list as the output recommendation information.

2. The method for recommending government service items according to claim 1, characterized in that, In the data collection step, the basic database includes an event database; The aforementioned database is used to store business processing item codes and reused business codes; The service item code is used to identify items within the same region; The reused service code is used to identify the same item across regions.

3. The method for recommending government service items according to claim 1, characterized in that, In the data processing step, the number of the aforementioned historical similar case records exceeds the data volume threshold.

4. The method for recommending government service items according to claim 1, characterized in that, In the seasonal effect weighting step, the method for weighting the seasonal effect is as follows: Based on the current month and historical frequency of occurrence, the support score of the historical similar case records is obtained, and the support score and the corresponding historical similar case records are used as weighted historical similar case records. The historical occurrence frequency refers to the historical occurrence frequency of the historical similar case records.

5. The method for recommending government service items according to claim 4, characterized in that, The support level is obtained by weighting the current month and the historical frequency of occurrence.

6. The method for recommending government service items according to claim 2, characterized in that, In the step of generating the historical user recommendation list, the method for sequence pattern mining is as follows: Based on the aforementioned event database, a global event sequence is obtained; Using the items from the user's historical completed cases in the basic database as a prefix, a projection database is constructed based on the global item sequence; The frequent sequence patterns are recursively mined in the projection database until the minimum support threshold is met. All frequent sequence patterns that meet the minimum support threshold are considered frequent sequences.

7. The method for recommending government service items according to claim 1, characterized in that, The method also includes a cold start step for newly established or newly introduced matters. If the newly introduced matters match the aforementioned identity characteristics. Or, the semantic similarity between the newly introduced item and the recommended list exceeds a semantic threshold, and the recommended list is a list of recommendations output as recommendation information; Or the newly introduced matters comply with policy promotion rules. The newly introduced items will then be added to the recommended list.

8. A government service recommendation system, characterized in that, The system includes the following modules: The data acquisition module obtains the basic database and user identity. The data processing module obtains corresponding identity features based on the identity; and obtains augmented data conditions based on the identity features. Based on the augmented data conditions and the basic database, several historical similar case records are obtained; The seasonal effect weighting module applies seasonal effect weighting to the aforementioned historical similar case records to obtain corresponding weighted historical similar case records. The recommendation strategy selection module, if the user has a history of processed cases, then performs the historical user recommendation list generation module; otherwise, it performs the new user recommendation list generation module. The historical user recommendation list generation module performs sequence pattern mining on the several weighted historical similar case records to obtain frequent sequences; Based on the historical case records and the frequent sequences, a recommendation list is obtained as the recommendation information output. The new user recommendation list generation module obtains several processing frequencies and several confidence levels based on several weighted historical similar case records; it sorts the several weighted historical similar case records based on the several processing frequencies and several confidence levels, and outputs a recommendation list as recommendation information.

9. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the method of any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1-7.

Images