Intelligent charging service quality influence factor analysis method and system based on multi-source heterogeneous data fusion

Abstract

The invention relates to the field of intelligent charging service optimization, in particular to an intelligent charging service quality influence factor analysis method and system based on multi-source heterogeneous data fusion, and the method comprises the steps: sequentially carrying out the NLP emotion analysis and feature extraction of the multi-source heterogeneous data related to the charging service quality, and obtaining the standardized data; adopting a pre-trained net recommendation value analysis model to map the standardized data and a preset business theme to establish a mapping relationship; when the net recommendation value of the net recommendation value analysis model fluctuates, obtaining an intelligent charging service influence factor based on the mapping relation; nLP sentiment analysis and feature extraction are adopted for multi-source heterogeneous data to obtain standardized data, the heterogeneous data are unified, the pre-trained net recommendation value analysis model is adopted for conducting cross-source mapping on the standardized data and the preset service theme, the net recommendation value is monitored in real time, and the user experience is improved. And reversely obtaining the intelligent charging service influence factors based on the mapping relationship, and realizing the attribution traceability of the influence factors.

Description

Technical Field

[0001] This invention relates to the field of intelligent charging service optimization, specifically to a method and system for analyzing the influencing factors of intelligent charging service quality through multi-source heterogeneous data fusion. Background Technology

[0002] With the development of the new energy vehicle industry, consumer demand for charging infrastructure has surged. To better meet these demands and support corporate decision-making, it is necessary to strengthen the monitoring of charging facilities and the quality control of platform services. Currently, user charging feedback data processing and analysis relies on manual analysis of tabular data to identify key factors influencing user evaluations of charging service quality. However, charging service operations involve massive heterogeneous data sources (structured evaluations, unstructured text such as user reviews, work order notes, etc.), with inconsistent standards and definitions, leading to difficulties in cross-source alignment, a lack of unified indicator systems, and the inability to create a visual decision dashboard. Existing evaluation systems primarily rely on single-source or offline data analysis, lacking efficient data access and integrated processing capabilities for structured / unstructured data. Current technologies have shortcomings in unified access and governance methods for cross-source heterogeneous data, making it difficult to effectively correlate and integrate data from different sources, creating data silos and hindering holistic analysis. Furthermore, the lack of domain knowledge constraints specific to charging scenarios makes it difficult for existing technologies to deeply understand the complex relationship between user feedback and service quality. Insufficient understanding of charging-specific concepts, processes, and user behaviors can lead to analysis results that are detached from reality and fail to accurately identify problems. Furthermore, the inadequacy of cross-source topic mapping exacerbates the difficulty in effectively correlating and integrating data from different sources. For example, "slow charging" in user reviews cannot be automatically correlated with "low charging power" in device operation data, making it impossible to verify and explain the problem from multiple dimensions. These factors combined prevent existing technologies from fully and deeply understanding the complex relationship between user feedback and service quality, and from comprehensively grasping service quality issues at both macro and micro levels, thus affecting problem diagnosis and solution development. Ultimately, this insufficient understanding of complex relationships severely restricts real-time monitoring and decision support for user experience and platform services, resulting in a lack of comprehensive and accurate insights and a lack of basis for decision-making.

[0003] Therefore, there is an urgent need to design a service quality influencing factor analysis method to address the core pain points in current charging service operations, such as data barriers, limited analytical dimensions, and insufficient decision support. Summary of the Invention

[0004] To address the technical problems of data silos, limited analytical dimensions, and insufficient decision support in current charging service operations, this invention proposes a method for analyzing the influencing factors of charging service quality through multi-source heterogeneous data fusion, including: Standardized data are obtained by sequentially performing NLP sentiment analysis and feature extraction on multi-source heterogeneous data related to charging service quality. A pre-trained Net Recommender Value (NPV) analysis model is used to map the standardized data to preset business themes to establish a mapping relationship. When the net promoter score (NPS) of the NPS analysis model fluctuates, the influencing factors of smart charging services are obtained based on the mapping relationship.

[0005] Optionally, the step of sequentially performing NLP sentiment analysis and feature extraction on the multi-source heterogeneous data related to charging service quality to obtain standardized data includes: Semantic classification of multi-source heterogeneous data related to charging service quality is performed using word segmentation, keyword matching, or weakly supervised clustering methods. NLP sentiment analysis methods were used to perform sentiment analysis and label the semantically categorized data. Standardized data is obtained by extracting features from the labeled data.

[0006] Optionally, a feature construction representation learning method based on statistics and deep learning can be used to extract features from the calibrated data to obtain standardized data.

[0007] Optionally, before performing NLP sentiment analysis on multi-source heterogeneous data related to charging service quality, the following may also be included: Perform data cleaning, missing value handling, field unification, and outlier identification on multi-source heterogeneous data.

[0008] Optionally, the training process of the net recommender value analysis model is as follows: Historical multi-source heterogeneous data related to charging service quality were sequentially subjected to NLP sentiment analysis and feature extraction to obtain historical standardized data. Based on the classification of pre-defined business themes, the types of historical standardized data are manually labeled and combined with the historical standardized data to form a training set; The net recommender value analysis model is trained using the training set.

[0009] Optionally, the step of obtaining the influencing factors of smart charging services based on the mapping relationship includes: The influencing factors of the smart charging service are screened using aggregate statistics or pre-trained machine learning methods to obtain the target influencing factors of the smart charging service.

[0010] Optionally, the multi-source heterogeneous data related to charging service quality is obtained using a multi-source data automated acquisition component.

[0011] A second aspect of the present invention provides a system for analyzing factors affecting the quality of intelligent charging services through multi-source heterogeneous data fusion, comprising: The data processing layer is used to sequentially perform NLP sentiment analysis and feature extraction on multi-source heterogeneous data related to charging service quality to obtain standardized data. The Net Promoter Score (NPS) analysis layer is used to establish a mapping relationship between the standardized data and preset business themes using a pre-trained NPS analysis model. The application layer is used to obtain the influencing factors of smart charging services based on the mapping relationship when the net promoter value of the net promoter value analysis model fluctuates, and to visualize the results.

[0012] Optionally, the data processing layer sequentially performs NLP sentiment analysis and feature extraction on multi-source heterogeneous data related to charging service quality to obtain standardized data. The steps include: Semantic classification of multi-source heterogeneous data related to charging service quality is performed using word segmentation, keyword matching, or weakly supervised distance methods. NLP sentiment analysis methods were used to perform sentiment analysis and label the semantically categorized data. Standardized data is obtained by extracting features from the labeled data.

[0013] Optionally, the data processing layer uses a feature construction representation learning method based on statistics and deep learning to extract features from the calibrated data to obtain standardized data.

[0014] Optionally, before performing NLP sentiment analysis on the multi-source heterogeneous data related to charging service quality, the data processing layer further includes: Perform data cleaning, missing value handling, field unification, and outlier identification on multi-source heterogeneous data.

[0015] Optionally, the training process of the net recommender value (NPV) analysis model in the NPV analysis layer is as follows: Historical multi-source heterogeneous data related to charging service quality were sequentially subjected to NLP sentiment analysis and feature extraction to obtain historical standardized data. Based on the classification of pre-defined business themes, the types of historical standardized data are manually labeled and combined with the historical standardized data to form a training set; The net recommender value analysis model is trained using the training set.

[0016] Optionally, after the application layer obtains the influencing factors of smart charging service based on the mapping relationship, it includes: The influencing factors of the smart charging service are screened using aggregate statistics or pre-trained machine learning methods to obtain the target influencing factors of the smart charging service.

[0017] Optionally, the multi-source heterogeneous data related to charging service quality in the data processing layer is obtained using a multi-source data automated acquisition component.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a method and system for analyzing the influencing factors of smart charging service quality through multi-source heterogeneous data fusion. The method includes: sequentially performing NLP sentiment analysis and feature extraction on multi-source heterogeneous data related to charging service quality to obtain standardized data; using a pre-trained Net Promoter Score (NPS) analysis model to map the standardized data to preset business themes to establish a mapping relationship; and obtaining influencing factors of smart charging service based on the mapping relationship when the NPS of the NPS analysis model fluctuates. This invention unifies the access of multi-source heterogeneous data, uses NLP sentiment analysis and feature extraction to obtain standardized data, unifies the heterogeneous data, uses a pre-trained NPS analysis model to perform cross-source mapping between the standardized data and preset business themes, monitors the NPS in real time, and obtains influencing factors of smart charging service based on the mapping relationship, thus achieving attribution and tracing of the influencing factors. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method for analyzing the influencing factors of smart charging service quality based on multi-source heterogeneous data fusion proposed in this invention. Figure 2 The present invention proposes Figure 1 A detailed flowchart of step S1; Figure 3 This is a partial structural diagram of the intelligent charging service quality influencing factor analysis system based on multi-source heterogeneous data fusion proposed in this invention. Figure 4 This is a schematic diagram of the overall structure of the intelligent charging service quality influencing factor analysis system based on multi-source heterogeneous data fusion proposed in this invention. Figure 5 This is a schematic diagram of the workflow of the intelligent charging service quality influencing factor analysis system based on multi-source heterogeneous data fusion proposed in this invention. Detailed Implementation

[0020] This invention proposes a method and system for analyzing the influencing factors of intelligent charging service quality through multi-source heterogeneous data fusion. This addresses core pain points in current charging service operations, such as data silos, limited analytical dimensions, and insufficient decision support, forming a complete technological closed loop from data fusion and intelligent analysis to visualized decision-making. Firstly, a unified data access and processing module is involved to automatically collect and integrate multi-source heterogeneous data from different business systems, including unstructured feedback data generated by user apps (such as NPS ratings and comment text) and semi-structured data from maintenance work order systems (such as fault descriptions and processing records). Through preprocessing such as data cleaning, standardization, and correlation alignment, the "data silos" are broken down. This lays a solid data foundation for subsequent comprehensive analysis. Adopting a combination of rule-driven and data modeling approaches, it addresses challenges such as "multiple signatures for a single review" and "mixed positive and negative data" by comprehensively utilizing technologies like Natural Language Processing (NLP) and weakly supervised clustering to tackle massive amounts of unstructured text (such as user reviews and work order notes). Furthermore, a layered, loosely coupled architecture was used to develop an intelligent charging service quality influencing factor analysis system that integrates data collection, intelligent analysis, and visualization. This system provides operational decision-makers with intuitive, multi-dimensional service quality monitoring, NPS attribution analysis, and decision support through dynamic interactive dashboards, thereby empowering refined operations and continuous service quality improvement.

[0021] Example 1: A method for analyzing the influencing factors of smart charging service quality through multi-source heterogeneous data fusion, such as... Figure 1 As shown, it includes the following steps S1 to S3.

[0022] S1: Perform NLP sentiment analysis and feature extraction sequentially on multi-source heterogeneous data related to charging service quality to obtain standardized data.

[0023] The system integrates multi-source data acquisition components and API interfaces to collect various heterogeneous data related to charging service quality, such as charging service platform operation data and user feedback data, enabling unified access to data from different sources and in different formats.

[0024] In a further preferred embodiment, the multi-source heterogeneous data related to charging service quality are sequentially subjected to NLP sentiment analysis and feature extraction to obtain standardized data, such as... Figure 2 As shown, the steps include: Step 1: Use word segmentation, keyword matching, or weakly supervised clustering methods to perform semantic classification on multi-source heterogeneous data related to charging service quality; Step 2: Perform sentiment analysis and labeling on the semantically categorized data using NLP sentiment analysis methods; in particular, NLP sentiment analysis is used to analyze unstructured issues such as user reviews. NLP sentiment analysis methods can perform multi-label and sentiment polarity recognition, solving the problem of mixed positive and negative expressions in comment topics, and achieving fine-grained judgment of user feedback.

[0025] Step 3: Extract features from the labeled data to obtain standardized data, in preparation for subsequent steps.

[0026] In a further optimized scheme, a feature construction representation learning method based on statistics and deep learning is used to extract features from the calibrated data to obtain standardized data, in order to construct analysis dimensions. For example, the frequency of occurrence of user feedback tags, the ratio of positive to negative sentiment, and the net positive index are extracted as the main features.

[0027] In a further optimized approach, the following steps are included before performing NLP sentiment analysis on the multi-source heterogeneous data related to charging service quality: Perform data cleaning, missing value handling, field unification, and outlier identification on multi-source heterogeneous data.

[0028] For example, rule matching methods can be used to clean the data to eliminate noise; KNN interpolation and other methods can be used to complete the missing values ​​of the collected data; field mapping table matching and fuzzy string matching methods can be used to unify the fields of "topics" and "L3 indicators"; IQR (interquartile range) and other methods can be used to identify outliers and truncation methods can be used to process outliers.

[0029] S2: A pre-trained Net Recommender Value (NPV) analysis model is used to map the standardized data to preset business themes to establish a mapping relationship.

[0030] In a further preferred embodiment, the training process of the Net Promoter Score (NPS) analysis model is as follows: Historical multi-source heterogeneous data related to charging service quality were sequentially subjected to NLP sentiment analysis and feature extraction to obtain historical standardized data. Based on the classification of pre-defined business themes, the types of historical standardized data are manually labeled and combined with the historical standardized data to form a training set; The net recommender value analysis model is trained using the training set.

[0031] The NPS analysis model can establish a correspondence between standardized data and preset business themes (such as device availability, charging efficiency, payment experience, price and fees, environment and safety, customer service, and charging station navigation) to form a hierarchical indicator system.

[0032] S3: When the net promoter value of the net promoter value analysis model fluctuates, obtain the influencing factors of smart charging service based on the mapping relationship.

[0033] The model analyzes the fluctuations in the net promoter score (NPS) of the time-to-time model. When fluctuations occur, the model traces back to obtain the influencing factors of smart charging services based on the mapping relationship. More preferably, the model also uses aggregate statistics or pre-trained machine learning methods to screen the influencing factors of smart charging services to obtain the key factors affecting the NPS change, which are the influencing factors of the target smart charging service.

[0034] This invention focuses on multi-source heterogeneous data in smart charging services, aiming to establish an efficient and reliable automated data acquisition system. The research covers several key aspects: First, systematically identifying and analyzing the structure, format, and heterogeneous characteristics of various data sources in the platform, especially the currently predominantly tabular data; second, designing and implementing a flexibly scalable data access framework to adapt to the integration needs of more types of data in the future; third, constructing a unified data reading, parsing, and storage mechanism to ensure the integrity, consistency, and timeliness of data during its flow; and finally, establishing a preliminary data quality monitoring and governance scheme to lay a solid foundation for subsequent complex data processing and value mining.

[0035] This invention achieves efficient collection and integrated processing of structured and unstructured data through the integration of multi-source heterogeneous data automated acquisition components and API interfaces. Compared to existing technologies that rely on manual analysis of tabular data, this invention fully automates the processes of data acquisition, cleaning, and standardization, increasing processing efficiency by tens of times, significantly reducing labor costs, and enabling real-time processing of massive amounts of heterogeneous data. Through the processing and feature extraction of structured and unstructured evaluation feedback data, fine-grained semantic understanding of unstructured text such as user reviews is achieved. Combined with a cross-source topic mapping mechanism, a correspondence is established between user feedback, work order business classification, and standard evaluation dimensions (equipment availability, charging efficiency, payment experience, price and fees, environment and safety, customer service, and charging station navigation). This allows the system to verify and explain problems from multiple dimensions (e.g., automatically associating "slow charging" in user reviews with "low charging power" in equipment operation data), comprehensively and deeply understanding the complex relationship between user experience and service quality from both macro and micro perspectives, and accurately identifying key factors affecting service quality. By constructing a dynamic, interactive visualization dashboard, complex analysis results are presented in an intuitive and multi-dimensional manner, enabling real-time monitoring and in-depth diagnosis of NPS changes.

[0036] Example 2: Based on the same inventive concept, this invention also provides a system for analyzing factors affecting the quality of intelligent charging services through multi-source heterogeneous data fusion, such as... Figure 3As shown, it includes: The data processing layer is used to sequentially perform NLP sentiment analysis and feature extraction on multi-source heterogeneous data related to charging service quality to obtain standardized data. The Net Promoter Score (NPS) analysis layer is used to establish a mapping relationship between the standardized data and preset business themes using a pre-trained NPS analysis model. The application layer is used to obtain the influencing factors of smart charging services based on the mapping relationship when the net promoter value of the net promoter value analysis model fluctuates, and to visualize the results.

[0037] In a further preferred embodiment, the data processing layer sequentially performs NLP sentiment analysis and feature extraction on multi-source heterogeneous data related to charging service quality to obtain standardized data. The steps include: Semantic classification of multi-source heterogeneous data related to charging service quality is performed using word segmentation, keyword matching, or weakly supervised distance methods. NLP sentiment analysis methods were used to perform sentiment analysis and label the semantically categorized data. Standardized data is obtained by extracting features from the labeled data.

[0038] In a further preferred embodiment, the data processing layer employs a feature construction representation learning method based on statistics and deep learning to extract features from the calibrated data to obtain standardized data.

[0039] In a further preferred embodiment, the data processing layer further includes the following steps before performing NLP sentiment analysis on the multi-source heterogeneous data related to charging service quality: Perform data cleaning, missing value handling, field unification, and outlier identification on multi-source heterogeneous data.

[0040] In a further preferred embodiment, the training process of the net recommender value (NPV) analysis model in the NPV analysis layer is as follows: Historical multi-source heterogeneous data related to charging service quality were sequentially subjected to NLP sentiment analysis and feature extraction to obtain historical standardized data. Based on the classification of pre-defined business themes, the types of historical standardized data are manually labeled and combined with the historical standardized data to form a training set; The net recommender value analysis model is trained using the training set.

[0041] In a further preferred embodiment, after the application layer obtains the influencing factors of the smart charging service based on the mapping relationship, it includes: The influencing factors of the smart charging service are screened using aggregate statistics or pre-trained machine learning methods to obtain the target influencing factors of the smart charging service.

[0042] In a further preferred embodiment, the multi-source heterogeneous data related to charging service quality in the data processing layer is obtained using a multi-source data automated acquisition component.

[0043] In a further preferred embodiment, the system also includes a data source layer and a data acquisition layer. The data source layer is used to provide various multi-source heterogeneous data, and the data acquisition layer is used to acquire various multi-source heterogeneous data.

[0044] This system adopts a layered and modular architecture design, such as Figure 4 As shown, from bottom to top, the layers are the data source layer, data acquisition layer, data processing layer, NPS analysis layer, and application layer, aiming to achieve a complete closed loop from the aggregation and integration of multi-source heterogeneous data to upper-level intelligent analysis and decision support. Figure 5 As shown, the data source layer aggregates various heterogeneous data related to charging service quality, such as platform operation data and user feedback data. The data acquisition layer integrates multi-source data automated acquisition components and API interfaces to achieve unified access to data from different sources and in different formats. The data processing layer includes data cleaning and other preprocessing to eliminate noise, feature extraction to construct analysis dimensions, NLP sentiment analysis technology to parse unstructured text such as user reviews, and data standardization to convert all data into a unified standard, preparing it for upper-level analysis. The NPS analysis layer maps data to multiple core business themes such as evaluation indicators, customer service channels, charging services, and facility construction. Then, it performs real-time monitoring and in-depth cause diagnosis of NPS changes to accurately locate key factors affecting service quality. The application layer provides a visual dashboard for the decision support system and various intuitive displays such as NPS change indicator analysis. Users can quickly obtain a comprehensive assessment of service quality through this layer, understand the root causes of NPS fluctuations, and obtain accurate and quantifiable decision-making basis to drive continuous optimization of service quality.

[0045] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A method for analyzing the influencing factors of smart charging service quality through multi-source heterogeneous data fusion, characterized in that, include: Standardized data are obtained by sequentially performing NLP sentiment analysis and feature extraction on multi-source heterogeneous data related to charging service quality. A pre-trained Net Recommender Value (NPV) analysis model is used to map the standardized data to preset business themes to establish a mapping relationship. When the net promoter score (NPS) of the NPS analysis model fluctuates, the influencing factors of smart charging services are obtained based on the mapping relationship.

2. The method for analyzing the influencing factors of intelligent charging service quality based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The steps for obtaining standardized data by sequentially performing NLP sentiment analysis and feature extraction on multi-source heterogeneous data related to charging service quality include: Semantic classification of multi-source heterogeneous data related to charging service quality is performed using word segmentation, keyword matching, or weakly supervised clustering methods. NLP sentiment analysis methods were used to perform sentiment analysis and label the semantically categorized data. Standardized data is obtained by extracting features from the labeled data.

3. The method for analyzing the influencing factors of intelligent charging service quality based on multi-source heterogeneous data fusion according to claim 2, characterized in that, A feature construction representation learning method based on statistics and deep learning is used to extract features from the calibrated data to obtain standardized data.

4. The method for analyzing the influencing factors of smart charging service quality based on multi-source heterogeneous data fusion according to claim 1, characterized in that, Before performing NLP sentiment analysis on multi-source heterogeneous data related to charging service quality, the following steps are also required: Perform data cleaning, missing value handling, field unification, and outlier identification on multi-source heterogeneous data.

5. The method for analyzing the influencing factors of intelligent charging service quality based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The training process of the Net Promoter Score (NPS) analysis model is as follows: Historical multi-source heterogeneous data related to charging service quality were sequentially subjected to NLP sentiment analysis and feature extraction to obtain historical standardized data. Based on the classification of pre-defined business themes, the types of historical standardized data are manually labeled and combined with the historical standardized data to form a training set; The net recommender value analysis model is trained using the training set.

6. The method for analyzing the influencing factors of intelligent charging service quality based on multi-source heterogeneous data fusion according to claim 1, characterized in that, After obtaining the influencing factors of smart charging services based on the mapping relationship, the following is included: The influencing factors of the smart charging service are screened using aggregate statistics or pre-trained machine learning methods to obtain the target influencing factors of the smart charging service.

7. The method for analyzing the influencing factors of intelligent charging service quality based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The multi-source heterogeneous data related to charging service quality is obtained using a multi-source data automated acquisition component.

8. A system for analyzing factors influencing the quality of intelligent charging services through multi-source heterogeneous data fusion, characterized in that, include: The data processing layer is used to sequentially perform NLP sentiment analysis and feature extraction on multi-source heterogeneous data related to charging service quality to obtain standardized data. The Net Promoter Score (NPS) analysis layer is used to establish a mapping relationship between the standardized data and preset business themes using a pre-trained NPS analysis model. The application layer is used to obtain the influencing factors of smart charging services based on the mapping relationship when the net promoter value of the net promoter value analysis model fluctuates, and to visualize the results.

9. The intelligent charging service quality influencing factor analysis system according to claim 8, characterized in that, The data processing layer sequentially performs NLP sentiment analysis and feature extraction on multi-source heterogeneous data related to charging service quality to obtain standardized data. The steps include: Semantic classification of multi-source heterogeneous data related to charging service quality is performed using word segmentation, keyword matching, or weakly supervised distance methods. NLP sentiment analysis methods were used to perform sentiment analysis and label the semantically categorized data. Standardized data is obtained by extracting features from the labeled data.

10. The intelligent charging service quality influencing factor analysis system according to claim 8, characterized in that, Before performing NLP sentiment analysis on the multi-source heterogeneous data related to charging service quality, the data processing layer also includes: Perform data cleaning, missing value handling, field unification, and outlier identification on multi-source heterogeneous data.

Images