CN114912447B - Data processing method and equipment based on smart city - Google Patents

Abstract

The smart city-based data processing method, device and device provided by the present invention perform differential analysis on the service feedback phrase vector set vector set2 extracted from the second type of smart business processing records to obtain the feedback event theme elements of the digital city business feedback event , Change the service feedback phrase vector set vector set2 into the service feedback phrase vector set vector set3 based on the feedback event theme elements of the digital city business feedback events, so as to realize the service feedback description mining based on the feedback event theme elements, and use the feedback event theme elements based on the service feedback description mining. The service feedback phrase vector and the thermal service feedback phrase vector are used to analyze the feedback event state, which can improve the accuracy of the feedback event state analysis for different types of smart business processing records.

Description

Data processing method and equipment based on smart city

technical field

The present invention relates to the technical field of data processing, in particular, to a data processing method and device based on a smart city.

Background technique

Smart city is to use information and communication technology to sense, analyze, and integrate various key smart business information of the core system of city operation, so as to make intelligent responses to various needs including city services and smart business activities. Its essence is to use advanced information technology to realize the intelligent management and operation of the city, thereby creating a better life for the people in the city and promoting the harmonious and sustainable growth of the city.

At present, various business services of smart cities are constantly increasing, and some business service functions may be difficult to meet the actual needs of users. In response to such problems, related technologies focus on the analysis and processing of user feedback, but in a long time, such technologies are still not available. Grasp the pain points of user feedback, that is, it is difficult to accurately analyze user feedback events.

SUMMARY OF THE INVENTION

The present invention at least provides a data processing method and device based on a smart city.

The present invention provides a smart city-based data processing method applied to a smart city cloud device, the method comprising: obtaining a first type of smart business processing records and a second type of digital city business feedback events collected by a big data collection system Smart business processing records; perform service feedback description mining on the first type of smart business processing records to obtain a service feedback phrase vector set vector set1, and perform service feedback description mining on the second type of smart business processing records to obtain a service feedback phrase vector set vector set2; According to the service feedback phrase vector set vector set2, obtain the target feedback event theme elements of the digital city business feedback event; according to the target feedback event theme elements and the service feedback phrase vector set vector set2, obtain the service feedback phrase vector set vector set3; According to the service The feedback phrase vector set vector set1 and the service feedback phrase vector set vector set3 are used to obtain the feedback event analysis report of the digital city business feedback event.

In this way, in the embodiment of the present invention, the first type of smart business processing records and the second type of smart business processing records of the digital city business feedback events collected by the big data collection system are obtained; the first type of smart business processing records are described by service feedback Mining to obtain the service feedback phrase vector set vector set1, and performing service feedback description mining on the second type of smart business processing records to obtain the service feedback phrase vector set vector set2; and combining the service feedback phrase vector set vector set2 to obtain the describe the target feedback event theme elements of the digital city business feedback event; combine the target feedback event theme elements and the service feedback phrase vector set vector set2 to obtain a service feedback phrase vector set vector set3; combine the service feedback phrase vector set vector set1 and the service feedback phrase vector set vector set3 to obtain the feedback event analysis report of the digital city business feedback event. In this way, differentiated analysis is performed on the service feedback phrase vector set vector set2 extracted from the second type of smart business processing records, and the feedback event theme elements (such as target feedback event theme elements) of the digital city business feedback event are obtained. Based on the digital city business feedback event change the service feedback phrase vector set vector set2 into the service feedback phrase vector set vector set3, in order to realize the service feedback description mining based on the feedback event subject element, and use the service feedback phrase vector based on the feedback event subject element (such as service Feedback phrase vector set vector set3) and thermal service feedback phrase vector (such as service feedback phrase vector set vector set1) for feedback event status analysis, which can improve the accuracy of feedback event status analysis for different types of smart business processing records.

In some exemplary embodiments, according to the service feedback phrase vector set vector set1 and the service feedback phrase vector set vector set3, the feedback event analysis report of the digital city business feedback event is obtained, including: combining the service feedback phrase vector set vector set1 and the service feedback The phrase vector set vector set3 is combined to obtain the service feedback phrase vector set vector set4; the heat index of each service feedback phrase vector in the service feedback phrase vector set vector set4 is obtained, and the relationship network of the heat index is obtained; the service feedback phrase vector set vector set4 and the heat index are obtained. The index relationship network is weighted and fused to obtain the first fused phrase vector set; the first fused phrase vector set is subjected to differential analysis to obtain the feedback event analysis report of the digital city business feedback event.

In this way, when applied to this embodiment, the service feedback phrase vector in the service feedback phrase vector set vector set4 is multiplied by the members of the corresponding distribution label in the thermal index relation network, and the first fused phrase vector set can be obtained. The service feedback phrase vector in the vector set can enrich and completely reflect the personalized description fields of the key data window (key data set or data area) of the digital city business feedback event.

In some exemplary embodiments, obtaining the thermal index of each service feedback phrase vector in the service feedback phrase vector set vector set4, and obtaining a thermal index relationship network, includes: invoking a deep learning network to output each service feedback in the service feedback phrase vector set vector set4 The bias influence index value of the phrase vector, and the numerical list formed by the bias influence index value is determined as the first thermal index relation network; the service interaction data block is identified for the second type of smart service processing records, and the X value of the set data window is obtained. business interaction data blocks, X is greater than 1; based on the service feedback phrase vector set vector set4 and X business interaction data blocks, the second thermal index relation network is obtained; the first thermal index relation network and the second thermal index relation network are calculated and to get the thermal index relation network.

In this way, this embodiment is applied to call the deep learning network to output the first thermal index relationship network, and then create a second thermal index relationship network for the service feedback phrase vector set vector set4 through the idea of service interaction data block identification and influence index configuration. The learning network has a certain probability of ignoring the information of some key data windows when outputting the bias influence index value, and based on the pre-configured influence index, it further configures the important coefficients for the service feedback phrase vector in the service feedback phrase vector set vector set4, thereby improving the The defects that may be caused by the deep learning network, so as to ensure the all-round analysis of the key data window as much as possible, to ensure that the first fused phrase vector set can reflect the personalized description field of the key data window as completely as possible, and then can improve the analysis of the event state. accuracy.

In some exemplary embodiments, the X business interaction data blocks further include a relative distribution label and a business topic label of each business interaction data block, and based on the service feedback phrase vector set vector set4 and the X business interaction data blocks, the second The thermal index relationship network, including: for the area corresponding to each service feedback phrase vector in the service feedback phrase vector set vector set4, it is determined that the area corresponding to each service feedback phrase vector corresponds to the second type of smart business processing records multiple service interaction data units, and obtain the relative distribution labels of the multiple service interaction data units; for each service interaction data unit in the multiple service interaction data units, call the relative distribution label and X of each service interaction data unit The relative distribution labels of each service interaction data block, determine the correlation between each service interaction data unit and each service interaction data block in the X service interaction data blocks; based on each service interaction data unit and X service interaction data The correlation between each service interaction data block in the block and the service subject label of each service interaction data block, configure the important coefficients for each service interaction data unit, and obtain X benchmark important coefficients of each service interaction data unit ; Determine the set operation value of the X benchmark important coefficients as the influence index of each service interaction data unit; obtain the influence index of each service feedback phrase vector based on the influence index of each service interaction data unit, and assign each service The numerical list formed by the influence index of the feedback phrase vector is determined as the second thermal index relation network.

In this way, based on the pre-configured impact index, this embodiment further configures important coefficients for the service feedback phrase vectors in the service feedback phrase vector set vector set4 to obtain a second thermal index relationship network, thereby improving the defects that may be caused by the deep learning network. , thereby ensuring a comprehensive analysis of key data windows as much as possible.

In some exemplary embodiments, performing service feedback description mining on the first-type smart business processing records to obtain a service feedback phrase vector set vector set1, including: loading the first-type smart business processing records into a first network of a deep learning network The service feedback description mining layer is performed to obtain the service feedback phrase vector set vector set1; the service feedback description mining is performed on the second type of smart business processing records, and the service feedback phrase vector set vector set2 is obtained, including: loading the second type of smart business processing records Go to the second network layer of the deep learning network to perform service feedback description mining, and obtain the service feedback phrase vector set vector set2.

In this way, this embodiment is applied to call different network layers to perform service feedback description mining on the first type of smart business processing records and the second type of smart business processing records respectively. In view of the fact that the first network layer calls the example of the first type of smart business processing records. Obtained from prior knowledge debugging, the second network layer invokes the prior knowledge debugging of the second type of smart business processing record example, and then loads the first type of smart business processing record and the second type of smart business processing record to their corresponding networks. The layer performs service feedback description mining, and can extract service feedback phrase vectors with more diverse and complete personalized description fields.

In some exemplary embodiments, the deep learning network further includes a topic element identification layer, a plurality of third network layers, and an event state identification layer, wherein the second network layer, the topic element identification layer, and the plurality of third network layers are concatenated, The multiple third network layers are non-linked network layers, and each third network layer in the multiple third network layers corresponds to a different feedback event theme element, and the processing result of each third network layer is respectively the same as that of the first network layer. The processing results of the layers are combined and regarded as the raw material for the event state recognition layer.

In this way, applied to this embodiment, for digital city business feedback events with different feedback event theme elements, the corresponding third network layer can be called in the same deep learning network for processing, compared with digital city for different feedback event theme elements Business feedback events need to call different deep learning networks to analyze the status of feedback events, which can reduce the amount of processing operations and improve the anti-interference ability of the deep learning network. Each third network layer of the deep learning network can learn based on historical network variables. Thereby, the efficiency of network debugging is improved, and further, only one subject element identification layer is connected after the second network layer, which can reduce the running delay of the deep learning network.

In some exemplary embodiments, the service feedback phrase vectors in the service feedback phrase vector set vector set2 include at least one personalized description field among user feedback emotion, feedback item details, and feedback timing characteristics. According to the service feedback phrase vector Set the vector set2 to obtain the target feedback event subject elements of the digital city business feedback event, including: loading the service feedback phrase vector set vector set2 into the subject element identification layer, and through the subject element identification layer to no less than one personalized description field Perform differential analysis to obtain the subject elements of the target feedback event, including the event type; according to the subject elements of the target feedback event and the service feedback phrase vector set vector set2, obtain the service feedback phrase vector set vector set3, including: The third network layer corresponding to the event type is determined in the third network layer, and the service feedback phrase vector set vector set2 is loaded into the third network layer corresponding to the event type for service feedback description mining, and the service feedback phrase vector set vector is obtained. set3.

In this way, applied to this embodiment, the service feedback phrase vector set vector set2 is subjected to differential analysis through the subject element identification layer in the deep learning network, and the target feedback event subject elements (such as event types) of the digital city business feedback event are obtained, and then from the The third network layer corresponding to the subject element of the target feedback event is determined from a plurality of third network layers, and the third network layer is called to perform service feedback description mining on the service feedback phrase vector set vector set2, which can make the service feedback phrase vector set vector set3 carries the service feedback phrase vector bound to the subject element of the feedback event, which can improve the accuracy of the status analysis of the feedback event.

The present invention also provides a data processing device based on a smart city, which is applied to a smart city cloud device, and the device includes the following functional modules:

The processing record acquisition module is used for: obtaining the first type of smart business processing records and the second type of smart business processing records of the digital city business feedback events collected by the big data collection system; and performing service feedback on the first type of smart business processing records Description mining to obtain a service feedback phrase vector set vector set1, and performing service feedback description mining on the second type of smart business processing records to obtain a service feedback phrase vector set vector set2;

The feedback event analysis module is used for: combining the service feedback phrase vector set vector set2 to obtain the target feedback event theme elements of the digital city business feedback event; combining the target feedback event theme elements and the service feedback phrase vector set vector set2 to obtain the service feedback phrase vector set vector set3; combining the service feedback phrase vector set vector set1 and the service feedback phrase vector set vector set3 to obtain the feedback event analysis report of the digital city business feedback event.

The present invention also provides a smart city cloud device, including a processor and a memory; the processor is connected in communication with the memory, and the processor is configured to read and execute a computer program from the memory to implement the above method .

The present invention also provides a computer-readable storage medium on which a computer program is stored, and the computer program implements the above method when running.

For a description of the effects of the smart city cloud device and the computer-readable storage medium, please refer to the description of the above method.

In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below, and are described in detail as follows in conjunction with the accompanying drawings.

Description of drawings

In order to illustrate the technical solutions of the embodiments of the present invention more clearly, the following briefly introduces the accompanying drawings that are used in the embodiments, which are incorporated into the specification and constitute a part of the specification. The drawings illustrate embodiments consistent with the present invention, and together with the description, are used to illustrate the technical solutions of the present invention. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as a limitation of the scope. Other related figures are obtained from these figures.

FIG. 1 is a block diagram of a smart city cloud device according to an embodiment of the present invention.

FIG. 2 is a schematic flowchart of a data processing method based on a smart city according to an embodiment of the present invention.

FIG. 3 is a block diagram of a data processing apparatus based on a smart city according to an embodiment of the present invention.

Detailed ways

Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. Where the following description refers to the drawings, the same numerals in different drawings refer to the same or similar elements unless otherwise indicated. The implementations described in the illustrative examples below are not intended to represent all implementations consistent with the present invention. Rather, they are merely examples of apparatus and methods consistent with some aspects of the invention as recited in the appended claims.

FIG. 1 is a schematic structural diagram of a smart city cloud device 10 according to an embodiment of the present invention, including a processor 102 , a memory 104 , and a bus 106 . Among them, the memory 104 is used to store execution instructions, including memory and external memory. The memory can also be understood as internal memory, which is used to temporarily store the operation data in the processor 102 and the data exchanged with the external memory such as the hard disk. The memory exchanges data with the external memory. When the smart city cloud device 10 is running, the processor 102 and the memory 104 communicate through the bus 106, so that the processor 102 executes the smart city-based data processing method of the embodiment of the present invention.

Please refer to FIG. 2 . FIG. 2 is a schematic flowchart of a smart city-based data processing method provided by an embodiment of the present invention, which is applied to a smart city cloud device. The method exemplarily includes the content described in the following steps.

Step 21: Obtain the first-type smart business processing records and the second-type smart business processing records of the digital city business feedback events collected by the big data collection system.

In the embodiment of the present invention, the smart city cloud device can obtain the thermal smart business processing records and knowledge-based smart business processing records of the digital city business feedback events synchronously collected by the big data collection system in real time, and can also be obtained from related databases. The thermal smart business processing records and the knowledge-based smart business processing records of the digital city business feedback events collected simultaneously by the big data collection system. For example, on the basis of obtaining thermal smart business processing records and knowledge-based smart business processing records, the smart city cloud device generates the identification module based on the data extraction strategy and disassembles the first two smart business processing records respectively. Type 2 smart business processing records and Type 2 smart business processing records. Among them, the first type of smart business processing record is a part of the thermal smart business processing record, and the second type of smart business processing record is a part of the knowledge-based smart business processing record. The thermal smart business processing record carries thermal value information/attention information of business interaction (focusing on the numerical level), and the knowledge-based smart business processing record carries various characteristic information of business interaction (focusing on the characteristic knowledge level).

For example, obtaining the first-type smart business processing records and the second-type smart business processing records of the digital city business feedback events collected by the big data collection system may include the following contents.

PRO1: From the multiple sets of knowledge-based smart business processing records of digital city business feedback events recorded in the relevant database, determine the group with the highest content score as the target knowledge-based smart business processing record, wherein the multiple sets of knowledge-based smart business processing records The records are obtained by the uninterrupted analysis of the digital city business feedback events by the event detection module in the big data collection system. The smart city cloud device performs service feedback description mining on the feedback behaviors in the multiple sets of knowledge-based smart business processing records through the previously debugged record content processing network to obtain service feedback phrase vectors containing various feedback behavior characteristics, and then differentiates Analytical processing is performed to obtain the value of each group of knowledge-based smart business processing records in multiple sets of knowledge-based smart business processing records, and the group with the highest value is determined as the target knowledge-based smart business processing record.

PRO2: Perform content scoring and identification on the feedback behaviors in the multiple sets of thermal smart business processing records recorded in the relevant database, obtain the value of each set of thermal smart business processing records in the multiple sets of thermal smart business processing records, and determine The difference between the value degree and the value degree of the target knowledge-based smart business processing record. The smart city cloud device also extracts service feedback phrase vectors containing various feedback behavior characteristics through the record content processing network, and then performs differential analysis to obtain the value of each group of thermal smart business processing records.

PRO3: Take the group with the smallest difference between the value degree and the value degree of the target knowledge-based smart business processing record among the multiple sets of thermal smart business processing records as the pending thermal smart business processing record.

PRO4: Identify the feedback service interaction data blocks for the target knowledge-based smart service processing record and the pending thermal smart service processing record respectively, and obtain multiple (P) first service interaction data blocks corresponding to the target knowledge-based smart service processing record and Multiple (P) second service interaction data blocks corresponding to the pending thermal smart service processing records.

PRO5: Determine the common value between the P first service interaction data blocks and the P second service interaction data blocks. If the common value is less than the set threshold, the target knowledge-based smart service processing record and the pending thermal smart service are processed. The processing record is determined as the processing record binary group obtained by the big data collection system collecting the digital city business feedback event at the same time node, and the target knowledge-based smart business processing record and the pending thermal smart business processing record when the feedback business interaction data block is identified. In the identification module, the partial processing records are respectively disassembled from the target knowledge-based smart business processing records and the pending thermal smart business processing records, and the first-type smart business processing records and the second-type smart business processing records of the digital city business feedback events are obtained. Process records.

Among them, when the smart city cloud device needs to obtain processing records from the relevant database, it may be difficult for the smart city cloud device to distinguish which knowledge-based smart business processing records and thermal smart business processing records are at the same time node. From the collected information, for any digital city business feedback event, first determine the group with the highest content score from its multiple sets of knowledge-based smart business processing records, and then determine the group with the highest content score from all the thermal smart business processing records in the related database. The group with the smallest difference in value from this group of knowledge-based smart business processing records is regarded as the pending thermal smart business processing record, and then P feedback business interaction data blocks are selected to match the two smart business processing records with the business interaction data blocks. If the common value between the business interaction data blocks is less than the set threshold, it is considered that the pending thermal smart business processing record and the target knowledge-based smart business processing record are collected from the digital city business feedback events at the same time node. Under the task that the cloud device needs to obtain the processing record from the relevant database, it can improve the accuracy and reliability of the processing record pairing.

Step 22: Perform service feedback description mining on the first type of smart business processing records to obtain a service feedback phrase vector set vector set1, and perform service feedback description mining on the second type of smart business processing records to obtain a service feedback phrase vector set vector set2.

In the embodiment of the present invention, a deep learning network architecture is shown, and the deep learning network may include a first network layer (the network layer may be understood as a branch), a second network layer, and a subject element identification layer (which may be understood as a category classification branch), multiple third network layers and event state identification layers, where the first network layer is used to perform service feedback description mining (for example: service feedback feature extraction) on the loaded first type of smart business processing records, and obtain The service feedback phrase vector set vector set1, the second network layer is used to perform service feedback description mining on the loaded second-type smart business processing records, and obtain the service feedback phrase vector set vector set2, wherein, the service feedback phrase vector set vector set1 and The service feedback phrase vector set vector set2 contains the personalized description field of whether the salient features of the feedback behavior (for example: feedback method, feedback emotion, feedback demand, etc.) are highly concerned (more urgent), for example, the personalized description field can be It is no less than one of user feedback emotions, feedback item details and feedback timing characteristics. Further, both the first network layer and the second network layer can call multiple convolution kernels in series to perform service feedback description mining. The convolution kernel calls sliding averages of different scales corresponding to the coverage of different scales. It is understandable and can be achieved. The integration of service feedback phrase vectors at different levels, thus, the service feedback phrase vector set vector set1 and the service feedback phrase vector set vector set2 have more diverse personalized description fields.

Step 23: According to the service feedback phrase vector set vector set2, the target feedback event subject element of the digital city business feedback event is obtained.

In the embodiment of the present invention, the service feedback phrase vector set vector set2 is loaded into the theme element identification layer, so that no less than one personalized description field is subjected to differential analysis processing through the theme element identification layer to obtain the target feedback event theme element. For example, the subject element of the target feedback event can be event heat, event timeliness, event type, etc. The subject element identification layer takes the feedback event subject element as a reference during the debugging process. The service feedback phrase vector set vector set2 can identify the target feedback event theme elements of the digital city business feedback event, such as: what type of digital city business feedback event is, and what kind of event timeliness the digital city business feedback event corresponds to.

Step 24: Obtain the service feedback phrase vector set vector set3 according to the target feedback event subject element and the service feedback phrase vector set vector set2.

In the embodiment of the present invention, after obtaining the target feedback event theme element of the digital city service feedback event, the smart city cloud device can determine the third network layer corresponding to the target feedback event theme element from multiple third network layers For example, if the event type of the digital city business feedback event is cross-border e-commerce, the cross-border e-commerce network layer, remote office network layer, virtual reality network layer and other third network layers can be determined. In the cross-border e-commerce network layer, the service feedback phrase vector set vector set2 is loaded into the cross-border e-commerce network layer for service feedback description mining, and the service feedback phrase vector set vector set3 is obtained. Further, multiple third network layers can similarly call multiple convolution kernels in series to perform service feedback description mining, and each third network layer calls the specified feedback event subject element as a reference during the debugging process, thus, The service feedback phrase vector set vector set3 includes the service feedback phrase vectors bound to the subject element of the feedback event, which can improve the accuracy of the status analysis of the feedback event.

Step 25: Obtain a feedback event analysis report of the digital city business feedback event according to the service feedback phrase vector set vector set1 and the service feedback phrase vector set vectorset3.

In this embodiment of the present invention, the feedback event analysis report may include the feedback cycle of the digital city service, the feedback event status of the digital city service, the activity of the digital city service, etc., which are not limited herein.

In a possible embodiment, according to the service feedback phrase vector set vector set1 and the service feedback phrase vector set vector set3, the feedback event analysis report of the digital city business feedback event is obtained, and the content recorded in the following steps 251-254 can be obtained. accomplish.

Step 251: Combine the service feedback phrase vector set vector set1 with the service feedback phrase vector set vectorset3 to obtain the service feedback phrase vector set vector set4.

Step 252: Obtain the thermal index of each service feedback phrase vector in the service feedback phrase vector set vector set4, and obtain a thermal index relationship network.

Step 253: Perform weighted fusion of the service feedback phrase vector set vector set4 and the thermal index relationship network to obtain a first fused phrase vector set.

Step 254: Perform differential analysis on the first fused phrase vector set to obtain a feedback event analysis report of the digital city business feedback event.

In the embodiment of the present invention, the thermal index can be understood as focusing on each service feedback phrase vector in the service feedback phrase vector set vector set4. Performing differential analysis on the first fused phrase vector set may be understood as classifying the first fused phrase vector set.

In a possible embodiment, the thermal index of each service feedback phrase vector in the service feedback phrase vector set vector set4 is obtained to obtain a thermal index relationship network (thermal index feature map), which may include the following content.

First, call the deep learning network to output the bias influence index value (which can be understood as the attention coefficient) of each service feedback phrase vector in the service feedback phrase vector set vector set4, and determine the numerical list formed by the bias influence index value as the first thermal index people network.

In this embodiment of the present invention, the deep learning network may be a priori deep learning network, and it should be understood that the deep learning network may perform mining analysis based on an attention mechanism, for example, by determining the service feedback phrase vector recorded by related processing. The bias effect index value of the service feedback phrase vector.

Secondly, the smart city cloud device identifies the service interaction data blocks of the second type of smart service processing records, and obtains X service interaction data blocks of the set data window and the relative distribution labels (location information) and relative distribution labels of the X service interaction data blocks. Business hashtags (category information).

Further, the set data window may refer to data sets in different regions, and the X service interaction data blocks may refer to P service interaction data blocks in PRO4.

Then, based on the service feedback phrase vector set vector set4 and X business interaction data blocks, the smart city cloud device determines to obtain the second thermal index relationship network,

Finally, by summing the members of the corresponding distribution labels of the first thermal index relation network and the second thermal index relation network, the thermal index relation network can be obtained.

Based on the above content, in a possible embodiment, the second thermal index relationship network is obtained based on the service feedback phrase vector set vector set4 and X service interaction data blocks, which can be achieved by the content recorded in the following steps 31 to 35 .

Step 31. For the area corresponding to each service feedback phrase vector in the service feedback phrase vector set vector set4, determine the multiple services corresponding to the area corresponding to each service feedback phrase vector in the second type of smart business processing record. Interactive data units, and obtain relative distribution labels of multiple business interactive data units.

In this embodiment of the present invention, in view of the fact that the service feedback phrase vector set vector set4 is obtained by combining the service feedback phrase vector set vector set1 and the service feedback phrase vector set vector set3, then based on the convolution check, the initial business processing records are subjected to a sliding average ( Convolution) and downsampling, in some examples, for the area corresponding to any service feedback phrase vector in the service feedback phrase vector set vector set4, it can be determined in the second type of smart business processing records. Multiple service interaction data units, for example, the service feedback phrase vector in this area may be determined by the service feedback phrase vector of the multiple service interaction data units. Further, it can be obtained that the multiple service interaction data units are in the second The relative distribution label in the smart-like business processing record, for example, the relative distribution label of a certain business interaction data block data block_a is (H4, V1).

Step 32: For each service interaction data unit in the plurality of service interaction data units, call the relative distribution label of each service interaction data unit and the relative distribution labels of the X service interaction data blocks, and determine the relationship between each service interaction data unit and each service interaction data unit. The correlation between each service interaction data block in the X service interaction data blocks.

In the embodiment of the present invention, suppose the relative distribution label of a certain service interaction data block data block_b in X service interaction data blocks is (H5, V7), then the relationship between the service interaction data block data block_a and the service interaction data block datablock_b The degree of correlation can be the Pearson correlation coefficient. Therefore, the correlation between each service interaction data unit in the plurality of service interaction data units and each service interaction data block in the X service interaction data blocks can be determined.

Step 33, based on the correlation between each business interaction data unit and each business interaction data block in the X business interaction data blocks and the business subject label of each business interaction data block, carry out important information for each business interaction data unit. Coefficient configuration, to obtain X benchmark important coefficients of each service interaction data unit.

In this embodiment of the present invention, impact indexes are set for X service interaction data blocks in advance, and the influence indexes of X service interaction data blocks may be Impact index1, Impact index2, . . . , Impact indexX. For the service interaction in multiple service interaction data units The data block data block_a, if the correlation between it and a certain service interaction data block (for example, the service interaction data block data block_b) in the X service interaction data blocks is less than the set correlation judgment value, the service interaction The impact index of the data block data block_b is assigned to the service interaction data block data block_a. If the correlation between it and the service interaction data block data block_b is greater than or equal to the set correlation judgment value, then the service interaction data block data block_a will be processed. The important coefficients are configured as 0. For example, the same impact index can also be set for the service interaction data blocks of the same service subject tag. For example, the impact indexes of the Y service interaction data blocks recorded in the first partial record can be set to Impact index1, and the impact index of the Y business interaction data blocks recorded in the second partial record can be set to Impact index1. The impact index of E business interaction data blocks can be set to Impact index2, and the impact index of K business interaction data blocks recorded in the third partial record can be set to Impactindex3, etc., then X business interaction data blocks also carry X impact indexes , the difference is that the influence index of the service interaction data block of the same service subject tag is also the same. For the service interaction data block data block_a and the service interaction data block data block_b, when the correlation between the two is less than the set correlation judgment value When the impact index of the business interaction data block data block_b is also configured to the business interaction data block data block_a, when the correlation between the two is greater than or equal to the set correlation judgment value, the same is performed for the business interaction data block data block_a. The important coefficients are configured as 0. Based on the above two influence index configuration ideas, each service interaction data unit is configured with X influence indexes, and the X influence indexes are used as the benchmark important coefficients of each service interaction data unit.

Step 34: Determine the set operation value of the X reference important coefficients as the influence index of each service interaction data unit.

In this embodiment of the present invention, the reference importance coefficient may be understood as a reference importance coefficient/reference weight.

Step 35: Obtain the influence index of each service feedback phrase vector based on the influence index of each service interaction data unit, and determine the numerical list formed by the influence index of each service feedback phrase vector as the second thermal index relation network.

In the embodiment of the present invention, since each service feedback phrase vector in the service feedback phrase vector set vector set4 corresponds to a plurality of service interaction data units in the second type of smart service processing records, based on each service interaction data in step 34 The influence index of the unit can determine the influence index of each service feedback phrase vector. For example, the set operation value of the influence index of multiple service interaction data units can be used as the influence of the corresponding service feedback phrase vector in the service feedback phrase vector set vector set4. index, for example, the central tendency variable of the influence index of multiple service interaction data units can also be used as the influence index of the corresponding service feedback phrase vector in the service feedback phrase vector set vector set4. Therefore, the influence index of each service feedback phrase vector The resulting list of values is determined as a second thermal index relationship network.

In the embodiment of the present invention, the service feedback phrase vector in the service feedback phrase vector set vector set4 is multiplied by the members of the corresponding distribution label in the thermal index relationship network, so as to obtain the first fused phrase vector set, the first fused phrase The service feedback phrase vectors in the vector set can enrich and completely reflect the personalized description fields of the key data windows of the digital city business feedback events.

Applied to this embodiment, the deep learning network is called to output the first thermal index relationship network, and then the second thermal index relationship network is created for the service feedback phrase vector set vector set4 through the idea of service interaction data block identification and impact index configuration. In view of the deep learning network When outputting the bias influence index value, there is a certain probability of ignoring the information of some key data windows, and based on the pre-configured influence index, the important coefficients are further configured for the service feedback phrase vector in the service feedback phrase vector set vector set4, thereby improving deep learning. Defects that may be caused by the network, so as to ensure all-round analysis of key data windows as much as possible, to ensure that the first fused phrase vector set can reflect the personalized description fields of key data windows as completely as possible, thereby improving the accuracy of event state analysis .

In a possible embodiment, a feedback event analysis report of the digital city business feedback event is obtained according to the service feedback phrase vector set vector set1 and the service feedback phrase vector set vector set3, including steps 41-46.

Step 41: Obtain the thermal index of each service feedback phrase vector in the service feedback phrase vector set vector set1, and obtain the thermal index relation network Network_E.

In the embodiment of the present invention, referring to the idea of obtaining the thermal index relationship network in step 252, first call the deep learning network to output the offset influence index value of each service feedback phrase vector in the service feedback phrase vector set vector set1, and set the offset The numerical list formed by the influence index value is determined as the third thermal index relationship network, and the service interaction data blocks are identified for the first type of smart service processing records, and Y service interaction data blocks of the set data window (for example, P service interaction data blocks) are obtained. data block) and the relative distribution labels and business topic labels of the Y service interaction data blocks, for the area corresponding to each service feedback phrase vector in the service feedback phrase vector set vector set1, determine the location of each service feedback phrase vector. The corresponding area corresponds to multiple service interaction data units in the first type of smart service processing records, and obtains the relative distribution labels of the multiple service interaction data units. For each service interaction data unit in the multiple service interaction data units , call the relative distribution label of each business interaction data unit and the relative distribution labels of the Y business interaction data blocks to determine the correlation between each business interaction data unit and each business interaction data block in the Y business interaction data blocks , based on the correlation between each business interaction data unit and each business interaction data block in the Y business interaction data blocks and the business subject label of each business interaction data block, perform a task for each business interaction data unit. The important coefficient configuration is to obtain Y benchmark importance coefficients of each business interaction data unit, and the set operation value of the Y benchmark importance coefficients is determined as the influence index of each business interaction data unit, and the multiple business interaction data units The set operation value or central tendency variable of the impact index is determined as the impact index of each service feedback phrase vector in the service feedback phrase vector set vector set1, and the numerical list formed by the impact index of each service feedback phrase vector is determined as For the fourth thermal index relationship network, the third thermal index relationship network and the fourth thermal index relationship network are summed to obtain the thermal index relationship network Network_E.

Step 42: Obtain the thermal index of each service feedback phrase vector in the service feedback phrase vector set vector set3, and obtain the thermal index relation network Network_F.

In the embodiment of the present invention, first call the deep learning network to output the bias influence index value of each service feedback phrase vector in the service feedback phrase vector set vectorset3, and determine the numerical list formed by the bias influence index value as the fifth thermal index relation network , identify the service interaction data blocks for the second type of smart service processing records, and also obtain U service interaction data blocks (for example, P service interaction data blocks) of the set data window and the relative distribution labels of the U service interaction data blocks and the business topic label, for the area corresponding to each service feedback phrase vector in the service feedback phrase vector set vector set3, determine the number of areas corresponding to each service feedback phrase vector in the second type of smart business processing records. each service interaction data unit, and obtain the relative distribution labels of the multiple service interaction data units, and for each service interaction data unit in the multiple service interaction data units, call the relative distribution label of each service interaction data unit and U The relative distribution label of the service interaction data block, to determine the correlation between each service interaction data unit and each service interaction data block in the U service interaction data blocks, based on each service interaction data unit and the U service interaction data The correlation between each service interaction data block in the block and the service subject label of each service interaction data block, configure important coefficients for each service interaction data unit, and obtain U benchmarks for each service interaction data unit Important coefficient, the set operation value of the U benchmark important coefficients is determined as the influence index of each service interaction data unit, and the set operation value or the central tendency variable of the influence indexes of the plurality of service interaction data units is determined as the service The influence index of each service feedback phrase vector in the feedback phrase vector set vector set3, the numerical list formed by the influence index of each service feedback phrase vector is determined as the sixth thermal index relational network, and the fifth thermal index relational network and the The sixth thermal index relation network is summed to obtain the thermal index relation network Network_F.

Those skilled in the art can know that the scale of data content carried by the service interaction data block and the service interaction data unit may be consistent or inconsistent, and both the service interaction data block and the service interaction data unit may be understood as the smallest constituent unit in the corresponding record.

Step 43: Multiply the service feedback phrase vector set vector set1 by the thermal index relationship network Network_E to obtain a second fused phrase vector set.

Step 44: Multiply the service feedback phrase vector set vector set3 by the thermal index relationship network Network_F to obtain a third fused phrase vector set.

Step 45: Combine the second fused phrase vector set with the third fused phrase vector set to obtain a combined fused phrase vector set.

Step 46: Perform differential analysis on the combined and fused phrase vector set to obtain a feedback event analysis report of the digital city business feedback event.

Applied to this embodiment, another method for combining service feedback phrase vectors is invoked, for example, by invoking a deep learning network, business interaction data block identification and impact index configuration as the service feedback phrase vector set vector set1, and outputting the bias impact list list1, followed by the service feedback The phrase vector set vector set1 is multiplied by the thermal index relationship network Network_E, and the second fused phrase vector set that can reflect the personalized description fields of the key data window as completely as possible can also be obtained; call the deep learning network, business interaction data block identification and The impact index is configured as the service feedback phrase vector set vector set3, and the output bias impact list list2 is multiplied by the service feedback phrase vector set vector set3 and the thermal index relationship network Network_F, and a personalized description that can reflect the key data window as completely as possible can also be obtained. The third fused phrase vector set of the field; the second fused phrase vector set and the third fused phrase vector set are combined to fully integrate the personalized description field of the hot data window in the second type of smart business processing record and the first The personalized description field of the hotspot data window in the quasi-smart business processing record can also improve the accuracy of event status analysis.

In this way, in the embodiment of the present invention, the first type of smart business processing records and the second type of smart business processing records of the digital city business feedback events collected by the big data collection system are obtained; the first type of smart business processing records are described by service feedback Mining to obtain the service feedback phrase vector set vector set1, and performing service feedback description mining on the second type of smart business processing records to obtain the service feedback phrase vector set vector set2; and combining the service feedback phrase vector set vector set2 to obtain the describe the target feedback event theme elements of the digital city business feedback event; combine the target feedback event theme elements and the service feedback phrase vector set vector set2 to obtain a service feedback phrase vector set vector set3; combine the service feedback phrase vector set vector set1 and the service feedback phrase vector set vector set3 to obtain the feedback event analysis report of the digital city business feedback event. In this way, differentiated analysis is performed on the service feedback phrase vector set vector set2 extracted from the second type of smart business processing records, and the feedback event theme elements (such as target feedback event theme elements) of the digital city business feedback event are obtained. Based on the digital city business feedback event change the service feedback phrase vector set vector set2 into the service feedback phrase vector set vector set3, in order to realize the service feedback description mining based on the feedback event subject element, and use the service feedback phrase vector based on the feedback event subject element (such as service Feedback phrase vector set vector set3) and thermal service feedback phrase vector (such as service feedback phrase vector set vector set1) for feedback event status analysis, which can improve the accuracy of feedback event status analysis for different types of smart business processing records.

In addition, for digital city business feedback events with different feedback event theme elements, the embodiment of the present invention can call the corresponding third network layer in the same deep learning network for processing, compared with digital city business for different feedback event theme elements Feedback events need to call different deep learning networks to analyze the state of feedback events, which can reduce the amount of processing operations and improve the anti-interference ability of the deep learning network. Each third network layer of the deep learning network can learn based on historical network variables. The efficiency of network debugging is improved, and further, only one subject element recognition layer is connected after the second network layer, which can reduce the running delay of the deep learning network.

In a possible embodiment, an idea of another smart city-based data processing method provided by the embodiment of the present invention, further, the method may include the content recorded in the following steps 61-68.

Step 61: Obtain the first-type smart business processing records and the second-type smart business processing records of the digital city business feedback events collected by the big data collection system. Step 62: Perform service feedback description mining on the first type of smart business processing records to obtain a service feedback phrase vector set vector set1, and perform service feedback description mining on the second type of smart business processing records to obtain a service feedback phrase vector set vector set2. Step 63: Obtain the target feedback event subject element of the digital city business feedback event according to the service feedback phrase vector set vector set2. Step 64: Obtain the service feedback phrase vector set vectorset3 according to the target feedback event subject element and the service feedback phrase vector set vector set2. Step 65: Combine the service feedback phrase vector set vector set1 with the service feedback phrase vector set vector set3 to obtain the service feedback phrase vector set vector set4. Step 66: Obtain the thermal index of each service feedback phrase vector in the service feedback phrase vector set vector set4, and obtain a thermal index relationship network. Step 67: Perform weighted fusion of the service feedback phrase vector set vector set4 and the thermal index relationship network to obtain a first fused phrase vector set. Step 68: Perform differential analysis on the first fused phrase vector set to obtain a feedback event analysis report of the digital city business feedback event.

For some independent embodiments, after obtaining the feedback event analysis report of the digital city service feedback event, the method may further include the following content: the feedback event analysis report in response to the digital city service feedback event characterizes all the feedback events. If the digital city business feedback event is in a state of high attention, and the digital city business feedback event is service function upgrade feedback, obtain the user feedback text corresponding to the digital city business feedback event; determine the function upgrade details based on the user feedback text. ; Utilize the features of the function upgrade rules to upgrade the smart city service function.

In this embodiment of the present invention, the state of high attention may be determined according to the attention threshold, the attention threshold may be set to 0.6, and the value range of attention may be 0 to 1. If the attention of the digital city business feedback event is higher than 0.6, then It can indicate that the digital city business feedback event is in a state of high concern (more urgent and needs to be resolved in time). Based on this, the user feedback text can be determined through the relevant smart business processing records, and then the features of the function upgrade details can be obtained by feature mining, so as to accurately and pertinently realize the smart city service function upgrade to solve a series of business service problems reported by users. .

For some independent embodiments, determining the feature of the function upgrade specification based on the user feedback text may be implemented with AI technology, and an example may include the following content.

S1, load the user feedback text into the word vector extraction network layer in the natural language processing model, obtain the first feedback text word vector and the second feedback text word vector of the user feedback text generated by the word vector extraction network layer, Wherein, the word vector extraction network layer includes a plurality of word vector extraction nodes connected in sequence, and the first feedback text word vector is a word except the last node in the sequentially connected multiple word vector extraction nodes The feedback text word vector generated by the vector extraction node, the second feedback text word vector is the feedback text word vector generated by the last word vector extraction node in the sequentially connected multiple word vector extraction nodes.

S2. Load the second feedback text word vector into the preliminary mining network layer in the natural language processing model, and obtain the target upgrade detailed rules description text generated by the preliminary mining network layer, wherein the target upgrade detailed rules description text Description text for the upgrade rules corresponding to the target function upgrade rules phrase identified in the user feedback text.

S3. Load the first feedback text word vector, the second feedback text word vector, the third feedback text word vector, and the description text of the target upgrade rules into the deep mining network layer in the natural language processing model, Obtain the phrase type of the target function upgrade detailed phrase and the distribution position of the core vocabulary of the target function upgrade detailed phrase in the user feedback text generated by the deep mining network layer, wherein the third feedback text word The vector is the feedback text word vector generated by the word vector extraction node in the preliminary mining network layer according to an initialization vector, and the initialization vector is a vector obtained by initializing the second feedback text word vector.

In the embodiment of the present invention, the preliminary mining network layer corresponds to the initial identification function, and the deep mining network layer corresponds to the secondary identification function. The distribution position of the core vocabulary in the user feedback text, so as to define and record the relevant features of the target function upgrade detail phrase as completely and accurately as possible, and provide an accurate and credible decision-making basis for the service upgrade process.

On the above basis, please refer to FIG. 3, the present invention also provides a block diagram of a data processing device 30 based on a smart city, the device includes the following functional modules:

The processing record obtaining module 31 is configured to: obtain the first type of smart business processing records and the second type of smart business processing records of the digital city business feedback events collected by the big data collection system; and to serve the first type of smart business processing records Feedback description mining to obtain a service feedback phrase vector set vector set1, and performing service feedback description mining on the second type of smart business processing records to obtain a service feedback phrase vector set vector set2;

The feedback event analysis module 32 is configured to: combine the service feedback phrase vector set vector set2 to obtain the target feedback event theme element of the digital city business feedback event; combine the target feedback event theme element and the service feedback phrase vector Set vector set2 to obtain the service feedback phrase vector set vector set3; combine the service feedback phrase vector set vector set1 and the service feedback phrase vector set vector set3 to obtain the feedback event analysis report of the digital city business feedback event.

Further, a readable storage medium is also provided, on which a program is stored, and when the program is executed by a processor, the above-mentioned method is implemented.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, for the specific working process of the system and apparatus described in the above exemplary description, reference may be made to the corresponding process in the foregoing method embodiments, which will not be repeated here. In the several embodiments provided by the present invention, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. The apparatus embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored, or not implemented. On the other hand, the shown or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces, indirect coupling or communication connection of devices or units, which may be in electrical, mechanical or other forms.

Claims (6)

1. A data processing method based on a smart city is applied to smart city cloud equipment, and the method comprises the following steps:

acquiring a first type of intelligent service processing record and a second type of intelligent service processing record of the digital city service feedback event collected by a big data collection system; service feedback description mining is carried out on the first type of intelligent service processing records to obtain a service feedback phrase vector set vector 1, and service feedback description mining is carried out on the second type of intelligent service processing records to obtain a service feedback phrase vector set vector 2;

combining the service feedback phrase vector set vector 2 to obtain a target feedback event topic element of the digital city service feedback event; combining the target feedback event topic element and the service feedback phrase vector set vector 2 to obtain a service feedback phrase vector set vector 3; combining the service feedback phrase vector set vector 1 and the service feedback phrase vector set vector 3 to obtain a feedback event analysis report of the digital city service feedback event;

the method further comprises the following steps:

responding to a feedback event analysis report of the digital urban business feedback event to represent that the digital urban business feedback event is in a high-attention state, and if the digital urban business feedback event is service function upgrading feedback, acquiring a user feedback text corresponding to the digital urban business feedback event; determining a function upgrade rule feature based on the user feedback text; upgrading the smart city service function by using the function upgrading detailed rule feature;

determining a function upgrade rule feature based on the user feedback text, comprising:

loading a user feedback text to a word vector extraction network layer in a natural language processing model to obtain a first feedback text word vector and a second feedback text word vector of the user feedback text, which are generated by the word vector extraction network layer, wherein the word vector extraction network layer comprises a plurality of word vector extraction nodes which are connected in sequence, the first feedback text word vector is generated by the word vector extraction nodes except the last node in the plurality of word vector extraction nodes which are connected in sequence, and the second feedback text word vector is generated by the last word vector extraction node in the plurality of word vector extraction nodes which are connected in sequence;

loading the second feedback text word vector to a primary mining network layer in the natural language processing model to obtain a target upgrading rule description text generated by the primary mining network layer, wherein the target upgrading rule description text is an upgrading rule description text corresponding to a target function upgrading rule phrase identified in the user feedback text;

loading the first feedback text word vector, the second feedback text word vector, a third feedback text word vector and the target upgrading rule description text to a deep mining network layer in the natural language processing model to obtain a phrase type of the target function upgrading rule phrase generated by the deep mining network layer and a distribution position of a core vocabulary of the target function upgrading rule phrase in the user feedback text, wherein the third feedback text word vector is a feedback text word vector generated by a word vector extraction node in the preliminary mining network layer according to an initialization vector, and the initialization vector is a vector obtained by initializing the second feedback text word vector;

and combining the service feedback phrase vector set vector 1 and the service feedback phrase vector set vector 3 to obtain a feedback event analysis report of the digital city service feedback event, wherein the step comprises the following steps:

obtaining the thermal indexes of service feedback phrase vectors in a service feedback phrase vector set1 to obtain a thermal index relational Network _ E, wherein a deep learning Network is specifically invoked to output bias influence index values of the service feedback phrase vectors in a service feedback phrase vector set vector 1, a numerical list formed by the bias influence index values is determined as a third thermal index relational Network, service interaction data block identification is carried out on a first-class intelligent service processing record, Y service interaction data blocks of a set data window and relative distribution labels and service subject labels of the Y service interaction data blocks are obtained in the same way, for a region corresponding to each service feedback phrase vector in a service feedback phrase vector set1, a plurality of service interaction data units corresponding to the region corresponding to each service feedback phrase vector in the first-class intelligent service processing record are determined, obtaining relative distribution labels of the plurality of service interaction data units, calling the relative distribution label of each service interaction data unit and the relative distribution labels of Y service interaction data blocks for each service interaction data unit in the plurality of service interaction data units, determining the correlation degree between each service interaction data unit and each service interaction data block in the Y service interaction data blocks, configuring important coefficients for each service interaction data unit based on the correlation degree between each service interaction data unit and each service interaction data block in the Y service interaction data blocks and the service theme label of each service interaction data block, obtaining Y reference important coefficients of each service interaction data unit, and determining the set operation value of the Y reference important coefficients as the influence index of each service interaction data unit, determining the set operation values or centralized trend variables of the influence indexes of the service interaction data units as the influence indexes of each service feedback phrase vector in a service feedback phrase vector set vector 1, determining a numerical list formed by the influence indexes of each service feedback phrase vector as a fourth thermodynamic index relationship Network, and summing the third thermodynamic index relationship Network and the fourth thermodynamic index relationship Network to obtain a thermodynamic index relationship Network _ E;

obtaining the thermal indexes of service feedback phrase vectors in a service feedback phrase vector set3 to obtain a thermal index relational Network _ F, wherein a deep learning Network is specifically invoked to output bias influence index values of the service feedback phrase vectors in a service feedback phrase vector set vector 3, a numerical list formed by the bias influence index values is determined as a fifth thermal index relational Network, service interaction data block identification is carried out on a second type of intelligent service processing record, U service interaction data blocks of a set data window and relative distribution labels and service subject labels of the U service interaction data blocks are obtained in the same way, for a region corresponding to each service feedback phrase vector in a service feedback phrase vector set3, a plurality of service interaction data units corresponding to the region corresponding to each service feedback phrase vector in the second type of intelligent service processing record are determined, obtaining relative distribution labels of the plurality of service interaction data units, calling the relative distribution label of each service interaction data unit and the relative distribution label of U service interaction data blocks for each service interaction data unit in the plurality of service interaction data units, determining the correlation degree between each service interaction data unit and each service interaction data block in the U service interaction data blocks, configuring important coefficients for each service interaction data unit based on the correlation degree between each service interaction data unit and each service interaction data block in the U service interaction data blocks and the service theme label of each service interaction data block, obtaining U reference important coefficients of each service interaction data unit, and determining the set operation value of the U reference important coefficients as the influence index of each service interaction data unit, determining the set operation values or centralized trend variables of the influence indexes of the service interaction data units as the influence indexes of each service feedback phrase vector in a service feedback phrase vector set vector 3, determining a numerical list formed by the influence indexes of each service feedback phrase vector as a sixth thermal index relationship Network, and summing the fifth thermal index relationship Network and the sixth thermal index relationship Network to obtain a thermal index relationship Network _ F;

multiplying a service feedback phrase vector set vector 1 with a thermodynamic index relationship Network _ E to obtain a second fused phrase vector set;

multiplying a service feedback phrase vector set vector 3 with a thermodynamic index relationship Network _ F to obtain a third fused phrase vector set;

combining the second fused phrase vector set with the third fused phrase vector set to obtain a combined fused phrase vector set;

and carrying out differential analysis on the combined fused phrase vector set to obtain a feedback event analysis report of the digital city service feedback event.

2. The method of claim 1, wherein said mining service feedback description of said first type of intelligent business process record to obtain a service feedback phrase vector set vector 1 comprises: loading the first type of intelligent service processing records to a first network layer of a deep learning network for service feedback description mining to obtain a service feedback phrase vector set 1;

the mining of the service feedback description of the second type of intelligent service processing record to obtain a service feedback phrase vector set2 includes: and loading the second type of intelligent service processing record to a second network layer of the deep learning network for service feedback description mining to obtain the service feedback phrase vector set vector 2.

3. The method according to claim 2, wherein the deep learning network further comprises a topic element identification layer, a plurality of third network layers and an event state identification layer, wherein the second network layer, the topic element identification layer and the plurality of third network layers are cascaded, the plurality of third network layers are non-linked network layers, each third network layer in the plurality of third network layers corresponds to a different feedback event topic element, and the processing result of each third network layer is respectively combined with the processing result of the first network layer and is regarded as a raw material of the event state identification layer.

4. The method of claim 3, wherein the service feedback phrase vector in the service feedback phrase vector set vector 2 includes at least one personalized description field of user feedback emotion, feedback item details and feedback timing characteristics, and the target feedback event topic element of the digital city service feedback event is obtained by combining the service feedback phrase vector set vector 2, and the method includes:

loading the vector set2 of the service feedback phrase vector to the topic element identification layer, and performing differentiation analysis processing on the at least one personalized description field through the topic element identification layer to obtain the target feedback event topic element, wherein the target feedback event topic element comprises an event type;

combining the target feedback event topic element and the service feedback phrase vector set vector 2 to obtain a service feedback phrase vector set vector 3, including: and determining a third network layer corresponding to the event type from the plurality of third network layers, and loading the service feedback phrase vector set vector 2 to the third network layer corresponding to the event type for service feedback description mining to obtain the service feedback phrase vector set vector 3.

5. A smart city cloud device comprising a processor and a memory; the processor is connected in communication with the memory, and the processor is configured to read the computer program from the memory and execute the computer program to implement the method of any one of claims 1 to 4.

6. A computer-readable storage medium, on which a computer program is stored which, when executed, implements the method of any of claims 1-4.

Images