A smart decision-making dynamic data analysis and early warning platform

Abstract

This invention relates to the field of big data analytics, specifically to an intelligent decision-making dynamic data analysis and early warning platform applied to business data monitoring and management. It includes an intelligent analysis module, comprising a data acquisition submodule for collecting business interaction information, a data analysis submodule for preprocessing and analyzing the collected business interaction information, a business rule setting submodule for configuring effective communication business rules, and an effective communication determination submodule for determining whether a business interaction is effective based on the analysis results of the data analysis submodule, the effective communication business rules, and sentiment analysis. The data analysis submodule also statistically analyzes the effective communication rate and the distribution of effective communication duration. Based on the effective communication business rules, this invention accurately analyzes whether effective communication exists in business interactions and further analyzes the effective communication rate and the distribution of effective communication duration.

Description

Technical Field

[0001] This invention relates to the field of big data analytics technology, specifically to an intelligent decision-making dynamic data analysis and early warning platform. Background Technology

[0002] With the rapid development of technologies such as big data and artificial intelligence, various big data-based analysis platforms are emerging. In terms of business data monitoring and management, although some data processing and analysis platforms have also appeared, the shortcomings of existing technologies are: 1. They focus more on basic semantic recognition, usually including sentiment, speech rate, and sensitive words, and the analysis is relatively simple; 2. They usually only perform compliance analysis, but the results of compliance analysis cannot truly reflect the interaction results, or even if the communication is compliant, it does not mean that the communication result is effective. Therefore, there is a lack of direct analysis of the effectiveness of the interaction. Summary of the Invention

[0003] The present invention aims to provide an intelligent decision-making dynamic data analysis and early warning platform. Based on the configured business rules for effective communication, it can accurately analyze effective communication and further analyze the effective communication rate and the distribution of effective communication duration, thereby providing a more comprehensive analysis of effective communication.

[0004] An intelligent decision-making dynamic data analysis and early warning platform is applied to business data monitoring and management. It includes an intelligent analysis module, which includes a data acquisition submodule, a data analysis submodule, a business rule setting submodule, and an effective communication and judgment submodule.

[0005] The data acquisition submodule is used to collect business interaction information;

[0006] The data analysis submodule is used to preprocess and analyze the collected business interaction information to form analysis results;

[0007] The business rules setting submodule is used to configure business rules for effective communication;

[0008] The effective communication determination submodule is used to determine whether the current business interaction is effective communication based on the analysis results of the data analysis submodule, the business rules of effective communication, and the sentiment analysis results.

[0009] The data analysis submodule is also used to calculate the effective communication rate and the distribution of effective communication duration.

[0010] The beneficial effects of this invention are as follows: The data acquisition submodule collects business interaction information, which can include information from multiple channels such as WeChat, SMS, voice, enterprise IM, and email, thus achieving comprehensive information collection. The business rule setting submodule allows for the configuration of effective communication business rules. This enables flexible configuration of these rules based on different industry specifics, and the configured rules provide analytical basis for determining whether the current business interaction is effective, making the analysis more reliable. The effective communication judgment submodule not only matches the analysis results from the data analysis submodule with the effective communication business rules but also incorporates sentiment analysis results from the interaction process to comprehensively determine whether the current business interaction is effective, making the determination of effective communication more accurate and reliable, providing strong support for business decision-making and early warning. This invention can configure business rules and analyze effective communication based on different industries and sales scenarios, making it flexible and widely applicable.

[0011] In this invention, the data analysis submodule also statistically analyzes the effective communication rate and the distribution of effective communication duration. By statistically analyzing the effective communication rate of a single business interaction or multiple business interactions, it can provide strong support for accurately analyzing whether customer service / sales personnel can accurately grasp and understand customer intentions, effectively guide customers, and solve customer problems during customer interactions. At the same time, this invention also statistically analyzes the effective communication duration, which can be used to accurately analyze customer needs / interests during business interactions, whether customer service / sales responses are helpful to customers, and whether customer service / sales communication is professional.

[0012] In a preferred embodiment of the present invention, the data analysis submodule is used to perform intent recognition, identifying the core intent of each customer's speech; entity recognition, automatically identifying key business entities; and sentiment analysis, analyzing the emotional tendencies and intensity of both parties in the communication. The data analysis submodule is also used to automatically identify the communication type based on the intent recognition results, entity recognition results, and sentiment analysis results. The communication types include pre-sales consultation, problem-solving, and relationship maintenance, and the communication types are configurable.

[0013] Explanation: The business entity described in this invention can be a product model, price, date, location, problem type, etc.; the intent can be to complain about service delays, schedule a demonstration, seek technical support, etc.; the sentiment analysis can include positive, negative, and neutral situations.

[0014] The beneficial effects are as follows: Through the data analysis submodule of this invention, the intent of each business interaction is identified, key business entities are identified, and the emotions of both parties in the communication are analyzed. Furthermore, based on the intent identification results, entity identification results, and sentiment analysis results, the communication type is classified and identified, thereby generating a more comprehensive and valuable business interaction analysis report. When customer service / sales fails to accurately and promptly understand the customer's intent, precise real-time script suggestions can be provided. When the emotions of both parties in the communication tend to be negative, timely reminders can be given to avoid adverse effects. As for the identification of key business entities, it can accurately analyze which business entities are involved in the business interaction process and the order in which each business entity appears, thereby enabling a more accurate analysis of the communication content, communication trend, and whether customer service / sales has provided timely and accurate guidance.

[0015] In this invention, communication types can be preset, i.e., customized according to business needs and flexibly configured. By classifying each business interaction, it is not only convenient to categorize and store each business interaction for easy viewing and subsequent data analysis, but also to fully understand which types of communication a customer service / sales representative uses most often and which types are most effective. This allows analysis of which types of problems the customer service / sales representative is good at solving or consulting, thus providing strong support for fully evaluating the quality of the customer service / sales representative's business interactions.

[0016] A preferred embodiment of the present invention is that the business rules for effective communication include: this communication involves key business entities, this communication achieves a specific intention, this communication uses standard language or process, and the communication duration / rounds exceed a threshold.

[0017] The beneficial effects are as follows: This invention allows for flexible configuration of effective communication business rules based on different business scenarios. It also allows for the unified configuration of standard, basic effective communication business rules within an enterprise. Furthermore, based on different business segments or scenarios within the enterprise, managers can add other effective communication business rules adapted to specific business situations, thus providing a basis for effective communication analysis. The effective communication business rules in this invention include, but are not limited to, the inclusion of key business entities in the communication. For example, merely exchanging greetings and pleasantries without mentioning any key business entities does not constitute effective communication. Secondly, a specific intention must be achieved, such as reaching an agreement on price or resolving a customer's problem. Professional language or standard procedures should be used during communication. Finally, the duration or number of communication rounds is also considered. One-way communication without any response from the customer is also not considered effective communication. By setting these rules, a basis for determining effective communication can be established, ensuring more accurate analysis of effective communication.

[0018] In a preferred embodiment of the present invention, the data analysis submodule is further used to perform script analysis, analyzing whether standard scripts, taboo words, and sensitive words are used; response efficiency analysis, analyzing the customer service / sales initial response time, average response time, and customer waiting time; and communication process standardization analysis, analyzing whether standard service processes are followed.

[0019] The beneficial effects are as follows: This invention, through a data analysis submodule, further analyzes the language used in business interactions, thereby accurately determining whether customer service / sales personnel use standard language, prohibited words, and sensitive terms during these interactions. When prohibited or sensitive terms are used, timely warnings are issued. It also performs response efficiency analysis, including initial response time, average response time, and customer waiting time. This analysis allows for timely assessment of response timeliness, whether customers are kept waiting excessively, and whether the average response time is too long or appropriate throughout the overall business interaction. This provides feedback on the quality of customer service / sales interactions based on response efficiency. Furthermore, this invention analyzes whether the communication process is standardized, providing a comprehensive analysis of the communication quality of customer service / sales personnel during business interactions from multiple dimensions and perspectives, resulting in a more complete and accurate analysis.

[0020] A preferred embodiment of the present invention is that the intent recognition method includes:

[0021] Collect historical business interaction data, annotate each dialogue text with intent, and form an annotated dataset;

[0022] Preprocessing of text data includes word segmentation, stop word removal, punctuation filtering, and stemming.

[0023] The text is converted into numerical feature vectors using the bag-of-words model, TF-IDF, and word embedding methods.

[0024] Combine contextual features and business entity features as auxiliary inputs;

[0025] The labeled dataset is divided into training, validation, and test sets according to a certain ratio. The model is trained using the training set and the hyperparameters, including the learning rate and hidden layer dimension, are adjusted using the validation set.

[0026] We selected the Naive Bayes machine learning model, which is suitable for text classification.

[0027] The training process employs the cross-entropy loss function and uses the Adam optimizer for gradient descent.

[0028] Use accuracy, precision, and recall metrics to evaluate model performance;

[0029] For intent categories with poor recognition performance, optimization can be achieved through data augmentation or by increasing the amount of labeled data.

[0030] By establishing the above intent recognition classification model and evaluating its performance using accuracy, precision, and recall metrics, we can ensure accurate and efficient identification of the intent expressed by the user during the dialogue process. This invention can optimize intent categories with poor recognition performance by data augmentation or increasing the amount of labeled data, thereby ensuring a wide and accurate coverage of intent recognition.

[0031] A preferred embodiment of the present invention further includes a visualization module, which is used to visualize the evolution of the emotions and key intentions of customers and customer service / sales personnel over time in a single business interaction. The data analysis submodule is also used to provide real-time script suggestions or risk warnings to customer service / sales personnel when the customer's emotions change.

[0032] The beneficial effects are as follows: Through the visualization module, the emotions and key intentions of customers and customer service / sales personnel can be displayed in real time during business interactions. On the one hand, this allows customer service / sales personnel to realize when either or both parties have negative emotions and make timely adjustments or guidance, while also accurately grasping the customer's intentions. On the other hand, it also provides managers with effective monitoring, so that when negative emotions are detected in customer service / sales personnel, timely alerts can be issued, and when customer service / sales personnel fail to accurately grasp the customer's intentions, timely suggestions on the communication script can be provided.

[0033] A preferred embodiment of the present invention is that the data analysis submodule is also used for customer trend analysis, dynamically analyzing the changing trends of the status, needs, value and risks of a certain customer or a certain customer group.

[0034] The beneficial effects are as follows: The data analysis submodule of this invention further conducts customer trend analysis, thereby enabling a comprehensive and accurate understanding of the changing trends of the status, needs, value, and risks of a particular customer or customer group over time, providing strong support for upper-level decision-making and early warning.

[0035] A preferred embodiment of the present invention is that the data analysis submodule, based on the analyzed core data, including core intent, key business entities, emotional tendencies and intensity, communication types, effective communication status, and other auxiliary data, performs customer trend analysis. This customer trend analysis includes: demand / interest evolution analysis, analyzing the changing trends of customer consultation topics and the frequency of product / service mentions; relationship health / satisfaction trend analysis, analyzing the average emotional tendencies and fluctuations of customers in past communications, analyzing changes in the effective communication rate, and analyzing changes in the distribution of communication types; and value potential change trend analysis, assessing the changing trends of the customer's current and potential value.

[0036] The beneficial effects are as follows: This invention first analyzes the intent, key business entities, emotional tendencies and intensity, communication types, and effective communication of a single customer service / sales interaction. The core intent, key business entities, and emotional tendencies and intensity may appear once or multiple times in a complete interaction. Then, in multiple interactions between the customer service / sales representative and a customer, the above-mentioned analytical items (core intent, key business entities, emotional tendencies and intensity, communication types, and effective communication in each interaction) are correlated. This allows for analysis of customer trends from multiple dimensions, including needs / interest evolution analysis, analyzing changes in customer needs / interests over time and with each communication, and further predicting what other customer needs remain—that is, the trends in customer needs / interests—providing support for customer service / sales representatives to further guide customer needs. It also analyzes the changing trends in customer consultation topics and the frequency of product / service mentions, thereby identifying the customer's main focus. Relationship health / satisfaction trend analysis helps understand the status of the relationship established between customer service / sales and the customer after multiple communications, as well as future relationship trends. It analyzes the average emotional inclination and fluctuations of customers throughout the communication process, identifying when customers are more positive or negative when problems are unresolved, and at what times of day are they most cheerful, thus facilitating better communication. Analysis of changes in effective communication rate allows for real-time monitoring of effective communication, such as whether effective communication increases with the number of communications. Analysis of communication type distribution changes accurately reveals the communication type used in each communication and its distribution after multiple communications, identifying predominant communication types and their evolution. Value potential trend analysis assesses the changing trends of the customer's current and potential value, enabling the development of corresponding communication strategies and early warning strategies based on the customer's current and potential value.

[0037] A preferred embodiment of the present invention is that the data analysis submodule further includes an application duration analysis unit, which is used to analyze the applications used by the terminal within a certain period of time, the duration of use, the number of users, and the usage details.

[0038] The beneficial effects are: by analyzing the applications, usage duration, number of users, and usage details used by the terminal within a certain period of time, it is possible to determine how much time the terminal used for business interaction spends on applications related to this business and how much time it spends on other applications, thereby accurately analyzing whether the business time allocation and workload are reasonable.

[0039] A preferred embodiment of the present invention further includes a decision adjustment module, which is used to sort the conversion rates of ineffective interactions to effective interactions, effective interactions to potential interactions, and potential interaction objects to enter the cooperation pool, and dynamically adjust the customer relationship allocation rules according to the conversion rate sorting results when at least one of the three conversion rates is lower than a set conversion rate value.

[0040] The beneficial effects are as follows: Through the above-mentioned decision-making adjustments, this invention can effectively avoid the continuous occurrence of low conversion rates. When a customer service representative / salesperson has a persistently low conversion rate for one or more items, customer relationships can be reallocated to optimize resource allocation, which is more conducive to improving customer relationship conversion rates and enhancing overall performance. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the architecture of the intelligent decision-making dynamic data analysis and early warning platform of the present invention. Detailed Implementation

[0042] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described below are only for explaining the present invention and do not limit the scope of protection of the present invention.

[0043] The terms "first," "second," etc., used in the specification, claims, and embodiments of this application are used to distinguish similar objects, rather than to describe a specific order or sequence.

[0044] The present invention will be further described in detail below through preferred embodiments:

[0045] Example 1

[0046] As attached Figure 1 The intelligent decision-making dynamic data analysis and early warning platform shown is applied to business data monitoring and analysis.

[0047] An intelligent decision-making dynamic data analysis and early warning platform is applied to business data monitoring and management, including an intelligent analysis module, a risk control module, and an AI early warning module. The intelligent analysis module includes a data acquisition submodule, a data analysis submodule, a business rule setting submodule, and an effective communication judgment submodule. The risk control module includes a sensitive word setting submodule, a business interaction financial statistics submodule, a sensitive operation statistics submodule, and a sensitive word statistics submodule.

[0048] The data acquisition submodule is used to collect business interaction information. In this embodiment, the business interaction includes communication records from multiple channels such as WeChat (text and voice), telephone voice (which needs to be converted to text), enterprise IM, and email. The voice data is converted into text using ASR (speech recognition) technology.

[0049] The data analysis submodule is used to preprocess and analyze the collected business interaction information to form analysis results. In this embodiment, the preprocessing includes data cleaning and structuring. Data cleaning removes irrelevant characters, emoticons (or retains them for sentiment analysis), stop words, and noise information (such as system prompts converted to text). Structuring involves dividing and labeling the original dialogue flow according to session, turn, and speaker (customer service / sales vs. customer).

[0050] The business rules setting submodule is used to configure business rules for effective communication. The business rules for effective communication include: this communication contains key business entities (such as product name, quotation amount), this communication achieves a specific intention (such as: clarifying purchase intention, successful appointment), this communication uses standard language or process (such as: fully introducing product FAB, successfully handling objections), and the communication duration / rounds exceed a threshold (excluding invalid small talk).

[0051] The effective communication determination submodule is used to determine whether the current business interaction is effective communication based on the analysis results of the data analysis submodule, the business rules for effective communication, and the sentiment analysis results; the data analysis submodule is also used to calculate the effective communication rate and the distribution of effective communication duration.

[0052] In this embodiment, the data analysis submodule performs intent recognition through semantic understanding and key information extraction, identifying the core intent of each customer's statement, such as: inquiring about the price of product A, complaining about service delays, scheduling a demonstration, seeking technical support, expressing purchase intention, or indicating refusal / hesitation, and establishing an intent classification model.

[0053] The intent classification model is built based on text classification techniques in Natural Language Processing (NLP). The specific steps are as follows:

[0054] Data preparation and annotation:

[0055] We collect a large amount of historical business interaction data (such as customer service chat logs, emails, etc.), and business experts annotate each dialogue text with intent based on actual scenarios, forming an annotated dataset. Common intent categories include, but are not limited to: inquiring about product prices, complaining about service issues, scheduling a demonstration, seeking technical support, expressing purchase intention, and indicating refusal / hesitation.

[0056] Preprocessing of text data includes word segmentation, stop word removal, punctuation filtering, and stemming.

[0057] Feature engineering:

[0058] Text is converted into numerical feature vectors using the Bag of Words model, TF-IDF (Term Frequency-Inverse Document Frequency), and Word Embedding methods (such as Word2Vec and GloVe).

[0059] Combine contextual features (such as dialogue turn, speaker role) and business entity features (such as product name, price, etc.) as auxiliary inputs.

[0060] Model selection and training:

[0061] We selected the Naive Bayes machine learning model, which is suitable for text classification.

[0062] The labeled dataset is divided into training, validation, and test sets according to a certain ratio. The model is trained using the training set and the hyperparameters, including the learning rate and hidden layer dimension, are adjusted using the validation set.

[0063] During training, the cross-entropy loss function is used, and the Adam optimizer is used for gradient descent.

[0064] Model evaluation and optimization:

[0065] Use accuracy, precision, and recall metrics to evaluate model performance.

[0066] For intent categories with poor recognition performance, optimization can be achieved through data augmentation (such as back translation and synonym replacement) or by increasing the amount of labeled data.

[0067] After the model is deployed, it is periodically fine-tuned with new data to maintain its adaptability and accuracy.

[0068] Specific content:

[0069] The final model can perform real-time intent classification on the input business interaction text, output the probability distribution of its belonging to a predefined intent category, and take the category with the highest probability as the recognition result. For example, when a user enters "I want to know how much product A costs?", the model should recognize the intent as "inquire about product price".

[0070] Entity recognition automatically identifies key business entities, such as: product model, price, date, name, location, contract number, problem type, and competitor name.

[0071] Sentiment analysis analyzes the emotional tendencies and intensity of both parties in the communication, such as positive, negative, and neutral.

[0072] Topic modeling: Automatically discovers frequently discussed topics and hot issues.

[0073] Topic modeling is used to automatically discover frequently discussed topics and hot issues in business interactions, specifically employing the following methods:

[0074] Algorithm selection:

[0075] Unsupervised learning algorithms, such as LDA (Latent Dirichlet Allocation) or NMF (Non-Negative Matrix Factorization), are used to extract latent topics from text data.

[0076] Data preprocessing:

[0077] The collected business interaction text is cleaned (irrelevant characters, emoticons, etc.) and then processed by word segmentation and stop word removal.

[0078] Construct a dictionary and a document-term matrix as input to the model.

[0079] Model training and topic extraction:

[0080] Set the number of topics (K value), and select the optimal number of topics by perplexity or coherence score.

[0081] The document-term matrix is ​​decomposed using LDA or NMF algorithms to obtain the term distribution for each topic and the topic distribution for each document.

[0082] For each topic, extract its Top-N keywords (such as "price", "fault", "renewal") and manually summarize topic tags (such as "price inquiry", "fault complaint", "renewal discussion").

[0083] Theme evolution analysis:

[0084] Data is sliced ​​by time window (e.g., weekly or monthly) to perform topic modeling, analyze the evolution trend of topic prevalence and keywords over time, and identify emerging or fading topics.

[0085] Applications and outputs:

[0086] The results of thematic modeling are combined with business rules for communication type classification (such as pre-sales consultation and problem-solving) and customer demand evolution analysis.

[0087] It outputs visual topic word clouds, topic trend charts, etc., to help managers quickly grasp business hotspots and risk points.

[0088] The data analysis submodule is also used to automatically identify the communication type based on the intent recognition result, entity recognition result, and sentiment analysis result. The communication type includes pre-sales consultation, problem-solving, and relationship maintenance, and the communication type is configurable.

[0089] Among them, the pre-sales consultation type focuses on product information, pricing, and advantages comparison; the problem-solving type focuses on troubleshooting, complaint handling, and technical support; and the relationship maintenance type focuses on customer care, renewal reminders, and satisfaction follow-ups.

[0090] Furthermore, the data analysis submodule is also used for dialogue analysis, analyzing whether standard dialogue, taboo words, and sensitive words are used, and analyzing the effectiveness of the dialogue (whether the target intent is achieved); response efficiency analysis, analyzing the customer service / sales' initial response time, average response time, and customer waiting time; and communication process standardization analysis, analyzing whether standard service processes are followed, such as: opening remarks, needs confirmation, solution introduction, objection handling, closing, and closing remarks. The data analysis submodule is also used for key node identification, thereby marking turning points in communication, such as: objections arising, price negotiations beginning, agreement reached, and customer emotional turning points. Its significance lies in providing important node references for the communication trend analysis of a single business interaction, making the communication trend analysis more accurate and better reflecting the communication effect and quality. Furthermore, when customers are in negative emotional states, unresponsive, or other problems arise that are unfavorable to facilitating cooperation, if a turning point occurs in the interaction, by analyzing whether communication turning points exist in multiple business interactions of a particular customer service / sales representative, it is possible to analyze and judge the strength of the customer service / sales representative's crisis public relations, opportunity acquisition, or problem-solving abilities in business interactions.

[0091] Furthermore, the intelligent decision-making dynamic data analysis and early warning platform in this embodiment also includes a visualization module. The visualization module is used to visualize the evolution of the customer's and customer service / sales' emotions and key intentions over time in a single business interaction. For example, the customer's emotions may change from anger (complaint) to calmness (acceptance of explanation) to satisfaction (problem solved). The data analysis submodule is also used to provide real-time script suggestions or risk warnings to customer service / sales when the customer's emotions change.

[0092] This invention can generate detailed conversation analysis reports (including intent, sentiment, entities, type, validity, quality score, etc.).

[0093] It can provide real-time script suggestions or risk alerts for customer service / sales (e.g., prompting for escalation when strong customer dissatisfaction is detected).

[0094] It can provide basic indicators for work quality analysis (communication duration, number of effective communications, response speed, compliance of scripts, etc.).

[0095] It can provide key inputs for subsequent customer trend analysis (such as: the intent, results, and customer sentiment of this communication).

[0096] In this embodiment, the data analysis submodule is also used for customer trend analysis, dynamically analyzing the changing trends of the status, needs, value, and risks of a specific customer or customer group. Specifically, the data analysis submodule performs customer trend analysis based on the analyzed core data, including core intent, key business entities, emotional tendencies and intensity, communication types, effective communication status, and other auxiliary data. This customer trend analysis includes: needs / interest evolution analysis, analyzing the changing trends of customer consultation topics and the frequency of product / service mentions; relationship health / satisfaction trend analysis, analyzing the average emotional tendency and fluctuations of customers in past communications, analyzing changes in the effective communication rate, and analyzing changes in the distribution of communication types; and value potential change trend analysis, assessing the changing trends of the customer's current and potential value.

[0097] In this embodiment, the auxiliary data includes: basic customer information (industry, size, region), transaction records (purchase history, contract amount, payment status), and other interaction records (activity participation).

[0098] Customer trend analysis is based on all interaction records (especially chat analysis results) between a specific customer or customer group and the enterprise over a period of time, dynamically depicting the changing trends of their status, needs, value, and risks, and making predictions.

[0099] Specifically, the core analytical dimensions are:

[0100] Evolution of needs / interests:

[0101] Analyze the changing trends in customer inquiry topics and the frequency of product / service mentions. For example, a shift from inquiring about basic functions to inquiring about advanced features to inquiring about integration solutions indicates a deepening of needs.

[0102] Identify new points of interest or questions that customers may have.

[0103] Relationship health / satisfaction trends:

[0104] Analyze the average emotional tendency and fluctuations of the client in each communication (whether negative emotions occur frequently? whether there is continuous improvement?).

[0105] Analyze changes in the rate of effective communication (is it becoming increasingly difficult to achieve effective communication?).

[0106] Analyze changes in the distribution of conversation types (e.g., from mainly pre-sales consultation to an increase in problem-solving, which may indicate an increase in product usage issues or service demands).

[0107] Changes in value potential:

[0108] Based on the budget, decision-making chain information, project scale (entity identification), and transaction history revealed during the communication, assess the changing trends of the client's current and potential value.

[0109] Analyze the ranking changes in customer growth rankings.

[0110] Risk warning indicators:

[0111] Decreased activity: Communication frequency significantly reduced and response time slower.

[0112] Accumulation / escalation of negative emotions: Negative emotions arise during multiple communications, and their intensity increases or they are not effectively resolved.

[0113] Key decision-maker attitude changes: (If the decision-maker can be identified in the communication) Analyze the negative shifts in the key decision-maker's emotions / intent.

[0114] Increased mentions of competitors: Frequent mentions of competitor products or comparisons during communication.

[0115] Interruption of effective communication: Repeated attempts to achieve effective interaction (e.g., the customer only asks questions without answering them, or avoids key issues).

[0116] Visualize the follow-up status (number of customers followed up):

[0117] The system dynamically displays the number of customers under the sales / customer service team at different stages (e.g., new leads, initial contact, needs confirmation, solution quotation, negotiation, pending contract, transaction, after-sales service) and their changing trends.

[0118] Identify clients who have been stagnant for a long time (e.g., no progress for more than 30 days in the "proposal quotation" stage).

[0119] Customer growth ranking:

[0120] Establish a comprehensive scoring model (with configurable weights) to rank customers periodically (e.g., weekly / monthly). Scoring factors include:

[0121] Interactive growth: growth rate of communication frequency and growth rate of effective communication.

[0122] Deepening of demand: The complexity of consulting issues has increased, and the scope of product lines / services involved has expanded.

[0123] Value enhancement: Estimated project value increases and purchase likelihood score improves.

[0124] Improved Relationships: Increased Satisfaction / Affectiveness Ratings.

[0125] Predictive early warning:

[0126] Based on historical trends, machine learning models (such as time series forecasting and classification models) are used to predict customer churn risk, purchase probability, and upgrade sales opportunities.

[0127] Trigger multi-dimensional early warnings (such as: "High-value customer A's satisfaction continues to decline warning", "Customer B's activity level has decreased significantly and there is a risk of churn", "Customer C's needs have been upgraded and senior consultant intervention is required).

[0128] Guidance strategy:

[0129] For customers with high growth potential, develop more proactive follow-up strategies.

[0130] For high-risk clients, promptly initiate retention plans (such as proactive care, discounts, and senior management intervention).

[0131] Identify common trends among customer groups to optimize product, service, and marketing strategies.

[0132] Output data such as "High-Growth Potential Customer Ranking" and "Risk Customer Ranking" to guide resource allocation.

[0133] This invention can automatically identify and classify conversation types: based on semantic understanding, it accurately classifies communication (such as pre-sales / problem / maintenance), providing a structured foundation for subsequent analysis.

[0134] Effective communication based on business rules: Going beyond basic quality inspection, dynamically defining and identifying interactions that are truly valuable for achieving business goals.

[0135] Dynamic visualization of conversational trends: gain a deeper understanding of the micro-evolution of a single communication session.

[0136] Customer trend dynamic analysis model: Integrates communication content, behavior, and transaction data to depict the continuous change trajectory of customer status / needs / risks.

[0137] Customer growth ranking quantitative model: Dynamically ranks customers' growth potential by integrating multi-dimensional interactions and value indicators.

[0138] Predictive multi-dimensional early warning based on multi-source data (especially deep chat analysis): Proactively predict risks and opportunities by combining multi-dimensional indicators such as customer trends and work quality.

[0139] Status-based and visual management of customer numbers: Provides a macro view of sales / customer service workload and progress.

[0140] Multi-dimensional integration of work quality analysis: It not only looks at the quality of a single communication (such as comparison documents), but also at long-term customer results (trends, growth rankings) and process management (follow-up status).

[0141] This invention enhances the depth and timeliness of customer insights, enabling a shift from reactive to proactive predictive management. It optimizes sales / customer service resource allocation and work prioritization, significantly improving customer satisfaction, retention, and conversion rates. It provides multi-dimensional, data-driven assessments of team and individual performance.

[0142] Example 2

[0143] Based on Embodiment 1, this embodiment of the intelligent decision-making dynamic data analysis and early warning platform further includes a risk control module. The risk control module includes a sensitive word setting submodule, a business interaction financial statistics submodule, a sensitive operation statistics submodule, and a sensitive word statistics submodule. The sensitive word setting submodule is used to add or modify existing sensitive words, including sensitive word classification settings, sensitive word content settings, and sensitive word statistical range settings. The business interaction financial statistics submodule is used to count the amount and frequency of financial transactions received and sent by each person in business interactions. The sensitive operation statistics submodule is used to count the sensitive operation content and frequency of each person in business interactions, and supports queries by setting time periods and / or sensitive word types and / or keywords. The sensitive word statistics submodule is used to count the sensitive words and frequency of each person in business interactions, and supports queries by setting time periods and / or statistical ranges and / or personnel information and / or sensitive word classifications.

[0144] The intelligent decision-making dynamic data analysis and early warning platform also includes an AI early warning module. The risk control adopts a three-level early warning mechanism. The AI ​​early warning module is used to issue a level one sensitivity warning to the management terminal when sensitive words, sensitive operations, and / or financial receipts and payments are detected. The AI ​​early warning module is also used to perform AI analysis on the interaction content related to financial receipts and payments, the interaction content related to sensitive operations, and the interaction content related to sensitive words. When further analysis reveals that at least one of the above three types of interaction content is abnormal, a level two sensitivity warning is issued to the management terminal. The AI ​​early warning module is also used to issue a level three sensitivity warning to the management terminal when the number of times the number of sensitive words, sensitive operations, and / or financial receipts and payments exceeds a set threshold, and the number of times the analyzed interaction content is abnormal also exceeds a set threshold.

[0145] The AI ​​early warning module is used to issue a Level 1 sensitivity warning to the management terminal when sensitive words, sensitive operations, and / or financial transactions are detected. In this embodiment, the Level 1 sensitivity warning serves as a prompt, indicating that the current personnel have engaged in the aforementioned sensitive words, sensitive operations, and / or financial transactions during business interactions. However, the Level 1 sensitivity warning alone cannot accurately determine whether the current personnel have engaged in inappropriate behavior during the interaction. For example, sensitive operations include creating groups, sending business cards, transferring money, or sending and receiving red envelopes. However, creating groups and sending business cards are not inappropriate behaviors but rather key steps in establishing intent with customers, which are positive behaviors. Therefore, to further distinguish between inappropriate and legitimate behaviors in the Level 1 sensitivity warning, this embodiment also classifies the content of sensitive words and sensitive operations according to their appropriateness. After detecting sensitive words and / or sensitive operations, the AI ​​early warning module filters out inappropriate sensitive words and behaviors, as well as positive sensitive words and behaviors, according to their appropriateness classification, and issues different forms of Level 1 sensitivity warnings, such as issuing positive Level 1 sensitivity warnings or inappropriate Level 1 sensitivity warnings.

[0146] In business interactions, sensitive words and their operations are a key aspect of risk control. Therefore, this invention first needs to define what constitutes a sensitive word. A sensitive word setting submodule is used to define the classification, content, and statistical scope of sensitive words. For example, the statistical scope may include only customer service / sales personnel or also sensitive words sent by customers. Based on this, a sensitive operation statistics submodule tracks whether sensitive operations occur and their frequency in each business interaction. A sensitive word statistics submodule also tracks whether sensitive words appear in each business interaction, their content, and classification information. Finally, a business interaction financial statistics submodule tracks the amount and frequency of financial transactions sent and received by each person in business interactions, thereby promptly identifying financial risk vulnerabilities and implementing timely early warning and control measures.

[0147] The AI ​​early warning module is also used to perform AI analysis on interactive content related to financial receipts and payments, interactive content related to sensitive operations, and interactive content related to sensitive words. When further analysis reveals that at least one of the above three types of interactive content is abnormal, a level-two sensitive warning is issued to the management terminal. The AI ​​early warning module is also used to issue a level-three sensitive warning to the management terminal when the number of times the statistically counted sensitive words, sensitive operations, and / or financial receipts and payments exceeds a set threshold, and the number of times the analyzed interactive content is abnormal also exceeds a set threshold.

[0148] This invention employs a three-tiered early warning mechanism. First, when sensitive words, sensitive operations, and / or financial transactions are detected, a Level 1 sensitivity warning is issued to the management system. This Level 1 warning, because it only alerts when the aforementioned situations are detected, may indicate problems or risks, but could also be normal business communication; therefore, it is used. Based on this warning, management personnel can determine the existence of risks or problems by combining it with the monitored contextual communication information. Further control measures are only implemented if risks are identified. Alternatively, after issuing a Level 1 sensitivity warning, AI analysis can be initiated to analyze interactions related to financial transactions, interactions related to sensitive operations, and interactions related to sensitive words. If further analysis reveals anomalies in at least one of these three types of interactions, a Level 2 sensitivity warning is issued to the management system. The system uses AI to further analyze sensitive interactions, enabling more accurate identification of any anomalies or risks associated with sensitive operations, keywords, or financial transactions. A Level 2 alert is issued, indicating that the detected sensitive information represents an abnormal operation requiring control. The system can send a Level 2 alert to the administrator along with control recommendations. Furthermore, the AI ​​alert module also issues a Level 3 alert to the management when the number of counted sensitive words, operations, and / or financial transactions exceeds a set threshold, and the number of abnormal interactions also exceeds a set threshold. A Level 3 alert indicates frequent sensitive interactions by a customer service / sales representative, prompting the administrator to implement stricter controls or other measures to prevent more serious risks.

[0149] The AI ​​early warning module is also used to analyze the business interaction content of each person in a randomly selected time period. If there are any abnormalities in the analyzed business interaction content, an interaction abnormality warning will be issued to the management terminal.

[0150] In this embodiment, the data analysis submodule includes a growth analysis unit, used to perform statistical analysis of customer growth, task growth, and task completion growth by individual and department, respectively, and generate data reports. These reports can be viewed and displayed by department or individual and by set time period. Through the statistical analysis of customer growth, task growth, and task completion growth, performance and business trends can be quickly and effectively analyzed for evaluating business quality.

[0151] The data analysis submodule also includes an interaction analysis unit, which is used to analyze interaction frequency, interaction volume, interaction duration, interaction status, and number of participants. Through the above more specific and detailed interaction analysis, the specific situation, interaction effect, frequency and duration of communication with customers can be analyzed more accurately and comprehensively.

[0152] The data analysis submodule also includes an application duration analysis unit, which analyzes the applications used by the terminal within a certain time period, the duration of use, the number of users, and the usage details. By analyzing the applications used by the terminal within a certain time period, the duration of use, the number of users, and the usage details can be determined how much time the terminal used for business interaction typically spends on applications related to this business, and how much time it spends on other applications, thereby accurately analyzing whether the business time allocation and workload are reasonable.

[0153] The data analysis submodule also includes a trend analysis unit. This submodule analyzes the comparative trends of intentional interactions, effective interactions, and ineffective interactions, providing a clear understanding of the differences between these three interaction outcomes. The interaction analysis submodule statistically analyzes different interaction durations for both intentional and effective interactions. Furthermore, the trend analysis submodule compares the duration trends of different interaction durations for both intentional and effective interactions, accurately analyzing the duration comparison between intentional and effective interactions and the correlation between them. The trend analysis submodule also compares the interaction volume over time, as well as the comparative trends of active and passive interactions over time.

[0154] The data analysis submodule also includes a conversion analysis unit. This submodule analyzes the number, conversion rate, and time taken for ineffective interactions to convert into effective interactions; the number, conversion rate, and time taken for effective interactions to convert into potential interactions; and the number, conversion rate, and time taken for potential interactions to be converted into the cooperation pool. Through this conversion analysis, the entire communication process from initial communication with a customer to understanding their needs, grasping their intentions, and finally reaching a cooperation agreement can be analyzed. This allows for a comprehensive understanding of the entire lifecycle conversion of all customers within the enterprise, providing strong support for customer maintenance, customer demand prediction, and risk warning.

[0155] In this embodiment, the intelligent decision-making dynamic data analysis and early warning platform further includes a decision adjustment module. This module sorts the conversion rates of ineffective interactions to effective interactions, effective interactions to potential interactions, and potential interactions leading to entry into the cooperation pool. When at least one of these three conversion rates falls below a set value, the module dynamically adjusts the customer relationship allocation rules based on the ranking results. Through this decision adjustment, the invention effectively prevents persistently low conversion rates. When a customer service representative / salesperson experiences consistently low conversion rates for one or more items, customer relationship allocation is re-optimized, optimizing resource allocation and improving overall performance.

[0156] The preferred embodiments of this application have been described in detail above with reference to the accompanying drawings. Typical known structures and common knowledge techniques in the preferred embodiments have not been described in detail here. Those skilled in the art can improve and implement the technical solutions of this invention based on the guidance provided in these embodiments and their own capabilities. Some typical known structures, known methods or common knowledge techniques should not be obstacles for those skilled in the art to implement this application.

[0157] The scope of protection claimed in this application shall be determined by the contents of its claims, and the contents described in the invention description, specific embodiments and drawings shall be used to interpret the claims.

[0158] Within the scope of the technical concept of this application, several modifications can be made to the specific implementation of this application, and these modified implementations should also be considered within the protection scope of this application.

Claims

1. An intelligent decision dynamic data analysis early warning platform applied to business data monitoring and management, characterized in that: Intelligent analysis module, including data acquisition submodule, data analysis submodule, business rule setting submodule, effective communication judgment submodule; The data acquisition submodule is used for collecting business interaction information; The data analysis submodule is used for preprocessing and analyzing the collected business interaction information to form an analysis result; The business rule setting submodule is used for configuring the business rules of effective communication; The effective communication judgment submodule is used for judging whether the current business interaction is effective communication according to the analysis result of the data analysis submodule and the business rules of effective communication in combination with the emotional analysis result; The data analysis submodule is also used for counting the effective communication rate and the effective communication time length distribution; The data analysis submodule further includes a conversion analysis unit, the conversion analysis submodule is used for analyzing the number, conversion rate and time length of invalid interaction converted into effective interaction, the number, conversion rate and time length of effective interaction converted into intended interaction, and the number, conversion rate and time length of intended interaction objects converted into a cooperation pool, and further includes a decision adjustment module, the decision adjustment module is used for sorting the conversion rates of invalid interaction converted into effective interaction, effective interaction converted into intended interaction and intended interaction objects converted into a cooperation pool respectively, and dynamically adjusting the customer emotion allocation rules according to the conversion rate sorting result when at least one of the three conversion rates is lower than a set conversion rate value; The data analysis submodule is used for intent recognition to identify the core intent of each speech of the customer; Entity recognition, automatically identifying key business entities; emotional analysis, analyzing the emotional tendency and intensity of both parties in communication, the data analysis submodule is also used for automatically identifying the communication type according to the intent recognition result, entity recognition result and emotional analysis result, the communication type includes pre-sale consultation type, problem solving type and relationship maintenance type, and the communication type is configurable; The business rules of effective communication include that the current communication contains key business entities, the current communication achieves a specific intent, the current communication uses standard dialogues or processes, and the communication time length / turns exceeds a threshold value. 2.The intelligent decision dynamic data analysis and early warning platform according to claim 1, characterized in that: The data analysis submodule is also used for dialogue analysis to analyze whether standard dialogues, taboos and sensitive words are used, response efficiency analysis to analyze the first response time, average response time and customer waiting time of the customer service / sales, and communication process standardization analysis to analyze whether the standard service process is followed. 3.The intelligent decision dynamic data analysis and early warning platform according to claim 1, characterized in that, The intent recognition method includes: Collecting historical business interaction data, labeling the intent of each dialogue text to form a labeled data set; Text data preprocessing, including word segmentation, stop word removal, punctuation symbol filtering and stem extraction; Using the bag-of-words model, TF-IDF and word embedding method to convert text into a numerical feature vector; Combining context features and business entity features as auxiliary input; Dividing the labeled data set into a training set, a validation set and a test set in proportion, training the model using the training set, and adjusting the hyperparameters including learning rate and hidden layer dimension through the validation set; Selecting a naive Bayes machine learning model suitable for text classification; The cross-entropy loss function is used in the training process, and the Adam optimizer is used for gradient descent; The model performance is evaluated using accuracy, precision, and recall indicators; For intent categories with poor recognition effect, data augmentation or increasing the amount of labeled data is used for optimization. 4.The intelligent decision dynamic data analysis and early warning platform according to claim 1, characterized in that: The visual display module is also included for visualizing the evolution of customer and customer service / sales emotions and key intents over time in a single business interaction. 5.The intelligent decision dynamic data analysis and early warning platform according to claim 1, characterized in that: The data analysis submodule is also used to provide real-time script suggestions or risk warnings to customer service / sales when customer emotions change. 6.The intelligent decision dynamic data analysis and early warning platform according to claim 5, characterized in that: The data analysis submodule is also used for customer trend analysis to dynamically analyze the changing trends of the status, needs, value, and risk of a certain customer or customer group. 7.The intelligent decision dynamic data analysis and early warning platform according to claim 1, characterized in that: The data analysis submodule analyzes core data, including core intent, key business entity, emotional tendency and intensity, communication type, effective communication, and other auxiliary data, to perform customer trend analysis, which includes demand / interest evolution analysis to analyze the changing trends of customer consultation topics and product / service mention frequency; relationship health / satisfaction trend analysis to analyze the average emotional tendency and fluctuation of customers in previous communications, analyze the change of effective communication rate, and analyze the distribution change of communication types; and value potential trend analysis to evaluate the changing trends of current value and potential value of customers. The data analysis submodule also includes an application duration analysis unit for analyzing the applications used by the terminal, the usage duration, the number of users, and the usage details within a certain time period.

Images