A big data decision analysis method and system

Abstract

The application relates to the technical field of big data analysis, and discloses a big data decision analysis method and system. The method receives market signals from different sources, carries out preliminary classification, starts the collection process of auxiliary scene information, and obtains multi-dimensional auxiliary scene information. On the basis, the application calculates the real influence of the market signals based on the original strength of the market signals in the preliminary classification and the multi-dimensional auxiliary scene information. The method can effectively identify and quantify the actual role of different market signals, and avoids the judgment errors caused by the fixed weight of information sources in the traditional method.

Description

Technical Field

[0001] This invention relates to the field of big data analytics, and more specifically, to a big data decision analysis method and system. Background Technology

[0002] In today's rapidly changing business environment, enterprise-level big data analytics platforms are playing an increasingly important role in supporting business decision-making. However, these platforms often face challenges such as low data integration efficiency and inaccurate extraction of key information when processing data from different sources. Traditional decision-making methods often rely on pre-set, fixed rules to process massive amounts of information. This makes them susceptible to interference from irrelevant information when facing complex and ever-changing market conditions, and they struggle to capture the hidden relationships between various pieces of information, leading to delayed decisions or misjudgments.

[0003] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0004] This invention provides a big data decision analysis method and system, aiming to solve the problems of low data integration efficiency, inaccurate extraction of key information, and difficulty in effectively identifying contradictory signals and judging the true influence weight of information sources when dealing with multi-source heterogeneous data in the face of complex and ever-changing market conditions, which leads to decision lag, judgment errors, or chaotic decision recommendations.

[0005] The technical solution of this application is as follows:

[0006] Firstly, this application discloses a big data decision analysis method, including:

[0007] Receive market signals from different sources and classify them to obtain preliminary classified market signals;

[0008] Based on the market signals identified in this preliminary classification, the process of collecting supplementary scenario information is initiated.

[0009] Obtain multi-dimensional auxiliary scenario information related to market signals associated with this preliminary classification;

[0010] Based on the original strength of the market signal in this preliminary classification and the multi-dimensional auxiliary scenario information, the true influence of the market signal is calculated.

[0011] Based on this true influence, conflicting market signals are ranked, and the dominant market signals are identified.

[0012] Based on this dominant market signal and its actual influence, decision-making recommendations are generated.

[0013] This technical solution effectively integrates market signals from different sources and multi-dimensional auxiliary scenario information. By calculating the true impact of market signals, it identifies and prioritizes contradictory signals, ultimately generating stable and reasonable decision-making recommendations. This solves the problems of chaotic decision-making recommendations and difficulty in coping with complex market changes in existing technologies.

[0014] Furthermore, in some implementations, the true impact of the market signal is calculated based on the original strength of the market signal according to the initial classification and the multi-dimensional auxiliary scenario information, including:

[0015] Receive text content from social media platforms and perform preliminary sentiment recognition on the text content to obtain preliminary sentiment tendency values;

[0016] Query the historical content style profile of the publisher of this text content, and calculate the publisher's ironic tendency index based on the frequency of use of rhetorical questions, exaggerated words, and puns in the historical content style profile, as well as the proportion of specific emojis appearing in the comments section interaction triggered by the publisher's content.

[0017] Monitor and collect comment section interaction data related to the text content, count the specific emojis that frequently appear in the comment section interaction data, and identify keywords and phrases that contain questioning, rhetorical questions, puns, or contradict the surface meaning of the text content, and calculate the interaction sentiment bias index;

[0018] Within a preset period of time after the text content is published, continuously monitor whether the text content is used by other users for secondary creation, meme dissemination, or over-interpretation, and calculate the deconstruction dissemination index;

[0019] The initial sentiment tendency value, the publisher's ironic tendency index, the interactive sentiment deviation index, and the deconstruction propagation index are used as inputs, and the calibrated sentiment tendency is calculated through the sentiment polarity correction function.

[0020] Based on this calibrated sentiment bias, the influence weight of social media signals in the overall decision analysis is adjusted, and the true influence of the market signal is calculated.

[0021] This technical solution enables the calibration of initial sentiment biases by analyzing social media text content from multiple dimensions, including publisher style, comment section interaction, and content dissemination methods. This allows for a more accurate assessment of the true impact of social media signals and avoids decision-making biases caused by misjudgments based on superficial sentiment biases.

[0022] Furthermore, in some implementation schemes, decision recommendations are generated based on the dominant market signals and their actual influence, including:

[0023] Assess the potential sales benefits, potential risks of unsold inventory, and costs associated with increasing inventory levels.

[0024] Assess the potential cost savings, potential sales losses, and market share loss risks associated with reducing this inventory decision direction;

[0025] By comparing the potential sales revenue, the potential risk and cost of unsold inventory, the potential cost savings, the potential sales loss, and the risk of market share loss, and combining these with the preset business objectives, a risk-adjusted decision score is generated.

[0026] Based on the risk-adjusted decision score, decision recommendations are generated.

[0027] This technical solution enables the quantitative assessment of the potential benefits, risks, and costs of different decision-making directions, and allows for risk adjustment in conjunction with business objectives. This helps identify the dominant decision-making direction amidst conflicting signals, preventing drastic swings in decision-making and improving the stability and rationality of decisions.

[0028] Building upon the above, this application further proposes that the method of receiving market signals from different sources and classifying these market signals to obtain preliminarily classified market signals includes:

[0029] Receive raw market signals from different sources and perform preliminary classification of these raw market signals;

[0030] When receiving promotional information from competitors, analyze the differences in technical specifications of the promoted products, the product life cycle stage, and the competitors' market strategies.

[0031] Based on the differences in technical specifications, the product life cycle stage, and the market strategy, the initial classification of promotional signals is revised.

[0032] When receiving a trending index signal on social media, analyze the characteristics of the group of publishers of trending content, key nodes in the content dissemination path, and the speed of diffusion;

[0033] Perform deep semantic analysis on the text of this trending content to identify whether there is any non-literal intent within it;

[0034] Based on the characteristics of the publisher group, the key nodes in the content dissemination path, the diffusion speed, and the non-literal intent, the preliminary classification of the online discussion popularity index signal is revised.

[0035] Based on the revised classification results, preliminary market signals are obtained.

[0036] This technical solution enables in-depth analysis and correction of different types of market signals (such as competitor promotional information and social media network discussion popularity index signals), thereby obtaining a more accurate preliminary classification, laying the foundation for subsequent calculation of true influence, and avoiding decision-making bias caused by inaccurate preliminary classification.

[0037] As a technological improvement, this application also proposes to calculate the true impact of the market signal based on the original strength of the market signal according to the preliminary classification and the multi-dimensional auxiliary scenario information, including:

[0038] Monitor the data update frequency of this multi-dimensional auxiliary scenario information;

[0039] Track the predictive accuracy of this multi-dimensional auxiliary contextual information in historical decision-making events;

[0040] Based on the data update frequency and the prediction accuracy, calculate the real-time effectiveness score of the multi-dimensional auxiliary scenario information;

[0041] Based on the real-time effectiveness score, the weight parameters of the multi-dimensional auxiliary scenario information in the scenario weighting function are dynamically adjusted.

[0042] The original strength of the market signal in the initial classification and the adjusted weight parameters are applied to the scenario weighting function to calculate the true impact of the market signal.

[0043] This technical solution enables the monitoring and tracking of the real-time effectiveness of multi-dimensional auxiliary scenario information, and dynamically adjusts its weight in the scenario weighting function. This ensures that the auxiliary scenario information can more accurately reflect the current market situation and improves the accuracy of calculating the true influence of market signals.

[0044] Based on the above, this application further proposes that, according to the real-time effectiveness score, the weight parameters of the multi-dimensional auxiliary scenario information in the scenario weighting function be dynamically adjusted, including:

[0045] Evaluate the real-time effectiveness score of this multi-dimensional auxiliary scenario information;

[0046] When the real-time effectiveness score is lower than the preset negative impact threshold for multiple consecutive monitoring periods, or when the real-time effectiveness score shows a continuous and significant downward trend compared with the historical average performance, the multi-dimensional auxiliary scenario information is judged to be inefficient or have a negative impact.

[0047] Perform a weight suppression operation, which includes setting the weight parameter of the multi-dimensional auxiliary scenario information in the scenario weighting function to zero, or multiplying the weight parameter by a preset minimum attenuation factor;

[0048] Record the suppressed multidimensional auxiliary context information and the reasons for its suppression;

[0049] During subsequent monitoring periods, the real-time effectiveness score of the suppressed multi-dimensional auxiliary scenario information will be continuously tracked.

[0050] When the real-time effectiveness score of the suppressed multi-dimensional auxiliary scenario information returns to the normal range and remains so for a period of time, the weight parameters of the multi-dimensional auxiliary scenario information are gradually restored.

[0051] This technical solution enables the timely identification of inefficient or negative auxiliary scenario information through real-time effectiveness scoring and evaluation, and the suppression of its weights. This avoids interference from invalid or harmful information in decision analysis, and gradually restores the weights as the information regains its effectiveness, ensuring the flexibility and accuracy of decision-making.

[0052] Furthermore, in some implementation schemes, if the real-time effectiveness score is below a preset negative impact threshold for multiple consecutive monitoring periods, or if the real-time effectiveness score shows a sustained and significant downward trend compared to the historical average performance, then the multi-dimensional auxiliary scenario information is determined to be inefficient or have a negative impact, including:

[0053] Obtain the short-term fluctuation range and long-term trend of the real-time effectiveness score;

[0054] Obtain a real-time effectiveness score for scenario information of the same type as this multi-dimensional auxiliary scenario information;

[0055] When the short-term fluctuation exceeds the preset short-term fluctuation threshold, the decrease in the real-time effectiveness score is judged as short-term market noise.

[0056] When the long-term trend indicates a continuous decline in the real-time effectiveness score, the decline in the real-time effectiveness score is judged to be a long-term trend.

[0057] When the real-time effectiveness score is lower than the average score of similar scenario information, and the average score of similar scenario information remains stable, it is judged that the multi-dimensional auxiliary scenario information is inefficient or has a negative impact.

[0058] Based on the judgment of short-term market noise, the judgment of long-term trend, and the judgment of the average score of similar scenario information, it is determined whether the multi-dimensional auxiliary scenario information is inefficient or has a negative impact.

[0059] This technical solution enables a more accurate distinction between short-term market noise and long-term downward trends by comprehensively analyzing the short-term fluctuations and long-term trends of real-time effectiveness scores and comparing them with similar scenario information. This helps avoid misjudgments and improves the accuracy of judging the inefficiency or negative impact of auxiliary scenario information.

[0060] Based on the above, this application further proposes that when the real-time effectiveness score is lower than a preset negative impact threshold for multiple consecutive monitoring periods, or when the real-time effectiveness score shows a continuous and significant downward trend compared to the historical average performance, the multi-dimensional auxiliary scenario information is judged to be inefficient or have a negative impact, including:

[0061] Obtain the type identifier of this multi-dimensional auxiliary scenario information;

[0062] Based on the type identifier, load the negative impact threshold and downward trend judgment time window corresponding to the type identifier from the preset type rule library;

[0063] The real-time effectiveness score is compared with the negative impact threshold of the load.

[0064] Within the loading downward trend judgment time window, analyze the changes in the real-time effectiveness score to determine whether the real-time effectiveness score has a continuous downward trend;

[0065] Based on the comparison and analysis results, it is determined whether the multi-dimensional auxiliary contextual information is inefficient or has a negative impact.

[0066] This technical solution enables the dynamic loading of personalized negative impact thresholds and downward trend judgment time windows based on the type of auxiliary scenario information, thereby achieving more refined judgment and improving the accuracy and adaptability of judging inefficient or negative impacts of different types of auxiliary scenario information.

[0067] As a further improvement, this application also proposes that when the real-time effectiveness score is lower than a preset negative impact threshold for multiple consecutive monitoring periods, or when the real-time effectiveness score shows a continuous and significant downward trend compared to the historical average performance, the multi-dimensional auxiliary scenario information is judged to be inefficient or have a negative impact, including:

[0068] Obtain the average level and fluctuation range of this multi-dimensional auxiliary scenario information in the period before the event occurs;

[0069] Retrieve the type, scope of impact, and estimated duration of currently occurring uncontrollable events;

[0070] Based on the type of the uncontrollable event, load the impact pattern of that type on different scenario information;

[0071] Compare the actual performance of this multi-dimensional auxiliary contextual information during the event with the impact pattern of its loading;

[0072] Determine whether the magnitude and duration of the score decline in the multi-dimensional auxiliary scenario information are consistent with the short-term impact characteristics predicted by the impact pattern, and whether the score decline exceeds the average level and fluctuation range before the event occurred;

[0073] Based on the assessment results, the long-term strategic value of this multi-dimensional auxiliary scenario information will be re-evaluated.

[0074] This technical solution enables a deeper analysis of the decline in scores of auxiliary scenario information by combining the impact patterns of uncontrollable events, distinguishing between short-term shocks and long-term value declines, thereby avoiding the erroneous suppression of information with long-term strategic value due to short-term events and improving the strategic and forward-looking nature of decision-making.

[0075] Secondly, this application also discloses a big data decision analysis system, including:

[0076] The receiving end is used to receive market signals from different sources and classify the market signals to obtain preliminarily classified market signals.

[0077] The computing end is used to initiate the collection process of auxiliary scenario information for the market signal of the preliminary classification; and to acquire multi-dimensional auxiliary scenario information related to the market signal of the preliminary classification.

[0078] Based on the original strength of the market signal in this preliminary classification and the multi-dimensional auxiliary scenario information, the true influence of the market signal is calculated.

[0079] The identification end is used to sort conflicting market signals based on their true influence and identify the dominant market signals.

[0080] The recommendation side is used to generate decision-making recommendations based on the dominant market signal and its actual influence.

[0081] This technical solution, through a systematic modular design, enables the reception, processing, calculation of real influence, identification of contradictory signals, and generation of decision-making recommendations from market signals. This provides an efficient and accurate big data decision analysis platform, effectively addressing the problems of chaotic decision-making recommendations and difficulty in responding to complex market changes in existing technologies.

[0082] Beneficial effects

[0083] The big data decision analysis method and system disclosed in this application receive market signals from different sources and perform preliminary classification, thereby initiating the process of collecting auxiliary scenario information and obtaining multi-dimensional auxiliary scenario information.

[0084] Building upon this foundation, this application calculates the true impact of market signals based on the initial strength of the pre-classified market signals and multi-dimensional auxiliary contextual information. This method effectively identifies and quantifies the actual effects of different market signals, avoiding judgment errors caused by fixed information source weights in traditional methods. Subsequently, conflicting market signals are ranked according to their true impact, and the dominant market signals are identified. This solves the problem that existing technologies cannot effectively identify the nature of contradictions and the mutual influence between conflicting information when faced with strong, contradictory signals. Finally, based on the dominant market signals and their true impact, stable and reasonable decision-making recommendations are generated, effectively avoiding drastic fluctuations in decision outputs. This allows supply chain execution departments to follow established procedures, not only seizing potential sales peaks but also maintaining good supplier relationships.

[0085] In summary, this application significantly improves the accuracy and stability of big data decision analysis by introducing multi-dimensional auxiliary contextual information and a real influence calculation mechanism, overcoming the shortcomings of existing technologies such as low data integration efficiency, inaccurate extraction of key information, difficulty in capturing hidden information correlations, and decision lag or judgment errors. Attached Figure Description

[0086] Figure 1 This is a flowchart of a big data decision analysis method provided in an embodiment of the present invention;

[0087] Figure 2 This is a flowchart of a method for calculating the true influence of market signals provided by an embodiment of the present invention;

[0088] Figure 3 This is a schematic diagram of the structure of a big data decision analysis system provided in an embodiment of the present invention. Detailed Implementation

[0089] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0090] Reference Figure 1 , Figure 1 This is a flowchart of a big data decision analysis method provided in an embodiment of the present invention, including:

[0091] S11: Receive market signals from different sources and classify the market signals to obtain preliminarily classified market signals;

[0092] S12, for the market signals that have been initially classified, initiate the process of collecting auxiliary scenario information;

[0093] S13, Obtain multi-dimensional auxiliary scenario information related to the market signals in the preliminary classification;

[0094] S14. Based on the original strength of the market signals in the preliminary classification and multi-dimensional auxiliary scenario information, calculate the true influence of the market signals.

[0095] S15, based on actual influence, sort conflicting market signals and identify the dominant market signals;

[0096] S16 generates decision-making recommendations based on dominant market signals and their actual influence.

[0097] This application, by introducing multi-dimensional auxiliary contextual information and a real influence calculation mechanism, can effectively identify and process contradictory market signals, thereby generating more stable and accurate decision-making recommendations and avoiding decision-making delays and errors caused by improper information processing in traditional methods.

[0098] The term "market signal" as used in this application refers to data or information from various channels that reflects market dynamics and consumer behavior, such as social media network discussion popularity indices, competitor promotional information, news reports, and industry reports. "Auxiliary contextual information" refers to supplementary information related to the market signal that helps in understanding and evaluating its true meaning, such as the publisher's background, dissemination path, and historical data. "True influence" refers to the actual degree of impact of the market signal on decision-making after calibration with multi-dimensional contextual information; it considers factors such as the signal's original strength, credibility, and dissemination characteristics.

[0099] Specifically, the method of this application first receives market signals from different sources and then classifies these market signals to obtain preliminarily classified market signals. For example, a data interface module can be set up to receive raw market signals from social media platforms (such as Weibo, WeChat, Twitter, etc.), news media (such as Reuters, Xinhua News Agency, etc.), industry report databases (such as iResearch, IDC, etc.), and publicly available information from competitors (such as official websites, financial reports, etc.).

[0100] Subsequently, for the initially categorized market signals, the system initiates the process of collecting auxiliary contextual information. For example, upon identifying a market signal in a preliminary category, the system can automatically trigger an information collection module. This module collects relevant data from preset auxiliary information sources based on the type and content of the market signal. For instance, if the market signal is about product reviews on social media, the system can initiate the collection of information such as the reviewer's historical posting records, the number of users' attention weights, and activity levels; if the market signal is about a competitor's promotional activities, the system can initiate the collection of information such as the competitor's historical promotional strategies, product line layout, and market share.

[0101] Next, multi-dimensional auxiliary contextual information related to the initially categorized market signals is acquired. For example, after initiating the auxiliary contextual information collection process, the system can use web crawling technology to scrape data from public websites or obtain data from partner data providers via API interfaces. This multi-dimensional auxiliary contextual information may include, but is not limited to: background information of the market signal publisher (such as user profiles, historical behavior patterns, and credibility), the propagation path of the market signal (such as the number of reposts, comments, and propagation nodes), the time context of the market signal (such as whether it is during holidays or promotional periods), relevant macroeconomic data (such as GDP growth rate and consumer confidence index), and changes in industry policies. For example, for product review signals on social media, the system can acquire the publisher's social influence, the tendency of their past comments, the number of times the comment was reposted and commented on, and the sentiment distribution in the comment section.

[0102] Building upon this foundation, the true influence of market signals is calculated based on the initial strength of the pre-classified market signals and multi-dimensional auxiliary contextual information. For example, an influence calculation model can be established that uses the initial strength of the pre-classified market signals (such as the number of likes on social media posts or the number of reads on news reports) as basic input and multi-dimensional auxiliary contextual information as correction factors. For instance, if a social media post receives a large number of likes, but its publisher has a low historical credibility and there are many dissenting voices in the comments section, its true influence may be significantly reduced. Conversely, if a post has a moderate number of likes, but the publisher is an industry authority, and the content is widely shared and sparks positive discussion, its true influence may be amplified. This calculation process can be implemented using weighted averaging, machine learning models (such as support vector machines or neural networks), or expert systems.

[0103] Furthermore, based on their true influence, conflicting market signals are ranked, and the dominant market signal is identified. For example, when the system receives conflicting signals of "declining demand" and "surging demand" simultaneously, these signals can be quantitatively compared based on their previously calculated true influence.

[0104] Finally, decision-making recommendations are generated based on dominant market signals and their actual impact. For example, once a dominant market signal is identified, the system can generate corresponding decision-making recommendations from a pre-set decision rule base or decision-making model, based on the signal's type, content, and actual impact. For instance, if a dominant "demand surge" signal is identified with extremely high actual impact, the system might recommend "immediately placing a large order with the supplier and initiating an emergency production plan." These decision-making recommendations can be presented to decision-makers in the form of text, charts, or visual dashboards, and can be accompanied by detailed analysis reports and risk assessments.

[0105] The overall working principle of this application lies in deeply calibrating the single-intensity assessment of traditional market signals by introducing multi-dimensional auxiliary contextual information, thereby revealing the true meaning and potential impact of market signals. Traditional methods often focus only on the surface intensity of market signals, such as the number of likes on social media or the exposure of news reports, while ignoring the complex context behind these signals, such as the publisher's intentions, the authenticity of the dissemination, and its connection with the macro environment. This single-dimensional assessment is prone to misjudging market signals, especially when faced with contradictory signals, making it impossible to effectively distinguish their priority and authenticity.

[0106] refer to Figure 2 , Figure 2 This is a flowchart of a method for calculating the true influence of market signals provided in an embodiment of the present invention. S14 specifically includes:

[0107] S141, Receive text content from a social media platform, and perform preliminary sentiment recognition on the text content to obtain a preliminary sentiment tendency value;

[0108] S142, query the historical content style profile of the publisher of the text content, and calculate the irony tendency index of the publisher based on the frequency of use of rhetorical questions, exaggerated words, and puns in the historical content style profile and the proportion of specific emojis appearing in the comment section interaction triggered by the publisher's content.

[0109] S143, monitor and collect comment section interaction data related to the text content, count the specific emojis that frequently appear in the comment section interaction data, and identify keywords and phrases that contain questioning, rhetorical questions, puns or contradict the surface meaning of the text content, and calculate the interaction sentiment deviation index;

[0110] S144. After the text content is published, continuously monitor whether the text content is used for secondary creation, meme dissemination or over-interpretation by other users within a preset period of time, and calculate the deconstruction dissemination index.

[0111] S145, taking the preliminary sentiment tendency value, the publisher's ironic tendency index, the interactive sentiment deviation index, and the deconstruction propagation index as inputs, the calibrated sentiment tendency is calculated through the sentiment polarity correction function;

[0112] S146, Based on the calibrated sentiment tendency, adjust the influence weight of social media signals in the overall decision analysis, and calculate the true influence of the market signals.

[0113] Specifically, receiving text content from social media platforms refers to scraping text information posted by users from various social media platforms such as Weibo, WeChat, Douyin, and Xiaohongshu. Preliminary sentiment identification of this text content can be performed using Natural Language Processing (NLP) techniques, combined with sentiment dictionaries and machine learning models, to analyze the sentiment polarity (positive, negative, neutral) and intensity of the text, thereby obtaining an initial sentiment tendency value. The purpose is to provide a basic assessment for subsequent refined adjustments.

[0114] The querying of the publisher's historical content style profile refers to the system maintaining a database containing information such as the user's past posting habits, language style, and common expressions. This historical content style profile records the frequency of the publisher's use of rhetorical questions, exaggerated vocabulary, and puns in their historical content, as well as the proportion of specific emoticons (e.g., laughing-crying, eye-rolling) appearing in the comments section. By analyzing these characteristics, an ironic tendency index can be calculated, reflecting the degree to which the publisher tends to use non-literal meanings or ironic techniques in their expression. Its purpose is to assess the true intention of the publisher's expression.

[0115] In practical applications, the system monitors and collects comment section interaction data related to the text content, including collecting user behavior data such as comments, likes, and reposts. It also analyzes frequently occurring specific emojis in the comment section interaction data; for example, if a seemingly positive comment is accompanied by a large number of "dog head" or "funny" emojis, it may imply sarcasm. Furthermore, it identifies keywords and phrases containing questioning, rhetorical questions, puns, or those that contradict the surface meaning of the text content.

[0116] Furthermore, within a preset period after the text content is published, continuous monitoring is conducted to determine whether the text content is used for secondary creation, meme dissemination, or over-interpretation by other users. This means observing whether the content is widely adapted, imitated, or made into internet slang or emojis, or whether its meaning is distorted or exaggerated during dissemination. For example, a originally serious discussion may be trivialized, or a neutral event may be given an extreme interpretation. By analyzing these dissemination phenomena, a deconstruction dissemination index is calculated. This index reflects the degree to which the original information is deconstructed, reconstructed, or alienated during dissemination. Its purpose is to assess the stability and authenticity of market signals from the perspective of information dissemination.

[0117] The initial sentiment tendency value, the publisher's ironic tendency index, the interactive sentiment deviation index, and the deconstruction propagation index are used as inputs to calculate the calibrated sentiment tendency using a sentiment polarity correction function. This sentiment polarity correction function is a pre-defined algorithm model that can weight, adjust, or non-linearly map the initial sentiment tendency value based on the values ​​of the aforementioned indices to obtain a sentiment tendency value that more closely approximates the true intention and influence.

[0118] Therefore, based on the calibrated sentiment bias, the influence weight of social media signals in the overall decision analysis is adjusted, and the true influence of the market signals is calculated. The calibrated sentiment bias can more accurately reflect the actual market sentiment and potential impact of social media signals. Therefore, when calculating the true influence of market signals, the weight of the signal in the overall decision model can be adjusted according to this calibration value, so that it can play a more appropriate role in the decision analysis.

[0119] This application's solution overcomes the limitations of traditional sentiment recognition methods in handling complex contexts by analyzing social media signals from multiple dimensions and at a deeper level. Through this technical solution, the accuracy of assessing the true impact of complex market signals such as those from social media can be significantly improved. Especially when faced with information containing irony, metaphor, or ambiguity, this solution can effectively identify and correct biases in traditional sentiment analysis, avoiding erroneous decisions caused by misinterpreting market sentiment.

[0120] This application further proposes decision-making suggestions, including:

[0121] Assess the potential sales benefits, potential risks of unsold inventory, and costs associated with increasing inventory levels.

[0122] Assess the potential cost savings, potential sales losses, and market share loss risks associated with reducing the aforementioned inventory decision direction;

[0123] By comparing the potential sales revenue, the potential unsold inventory risk and cost, the potential cost savings, the potential sales loss, and the market share loss risk, and combining these with preset business objectives, a risk-adjusted decision score is generated.

[0124] Based on the risk-adjusted decision score, decision recommendations are generated.

[0125] Specifically, when evaluating the decision to increase inventory, it is necessary to comprehensively consider the additional sales revenue that may be realized under scenarios of increased market demand or hot product sales, as well as the potential losses from unsold inventory due to inventory backlog, product expiration, or market changes, and related costs such as storage and management. Among them, potential sales revenue refers to the additional revenue that can be obtained by increasing inventory to meet higher demand; potential unsold inventory risks and costs cover the capital tied up due to excess inventory, warehousing costs, depreciation losses, and possible price reductions and promotions.

[0126] Meanwhile, when assessing the direction of inventory reduction decisions, the focus is on analyzing the operational cost savings that can be achieved by optimizing inventory levels, such as reducing warehousing, insurance, and capital tied up costs. However, it is also necessary to consider the sales opportunities that may be missed due to insufficient inventory, i.e., potential sales losses, as well as the risk of customer churn and market share decline that may result from the inability to meet market demand in a timely manner.

[0127] Furthermore, when comparing the aforementioned potential benefits, risks, and costs, a comprehensive consideration is given to the company's pre-set business objectives. For example, business objectives might include maximizing profits, minimizing risks, increasing market share, or optimizing cash flow. By combining these quantitative indicators with business objectives, a risk-adjusted decision score can be generated. This score quantifies the overall advantages and disadvantages of different decision directions (increasing or decreasing inventory), thus providing a quantitative basis for subsequent decision-making.

[0128] Ultimately, based on the generated risk-adjusted decision scores, the system can produce specific decision recommendations. For example, if the risk-adjusted decision score for increasing inventory is significantly higher than that for decreasing inventory, then increasing inventory is recommended; conversely, if the scores are close, further data analysis or manual intervention may be necessary.

[0129] This application overcomes the limitations of simply ranking market signals based on their actual impact by introducing a quantitative assessment of the potential benefits, risks, and costs of specific decision-making directions (such as increasing or decreasing inventory). Through this technical solution, the application significantly improves the accuracy and practicality of big data decision analysis when dealing with conflicting market signals. Especially in critical business decisions such as inventory management, it no longer relies solely on the surface strength of market signals but delves into the specific business impacts and risks of different decision directions. Consequently, the generated decision recommendations not only consider the actual impact of market signals but also incorporate the company's own risk appetite and business objectives, making the recommendations more targeted, actionable, and robust. This helps companies more effectively mitigate risks and seize opportunities in a complex and volatile market environment, thereby optimizing resource allocation and improving overall operational efficiency and market competitiveness.

[0130] This application further proposes the steps of receiving market signals from different sources and classifying the market signals to obtain preliminarily classified market signals, including:

[0131] Receive raw market signals from different sources and perform preliminary classification of the raw market signals;

[0132] When receiving promotional information from competitors, analyze the differences in technical specifications of the promoted products, the product life cycle stage, and the competitors' market strategies.

[0133] Based on the aforementioned differences in technical specifications, the aforementioned product life cycle stage, and the aforementioned market strategy, the initial classification of promotional signals is revised;

[0134] When receiving a trending index signal on social media, analyze the characteristics of the group of publishers of trending content, key nodes in the content dissemination path, and the speed of diffusion;

[0135] Perform deep semantic analysis on the text of the hot topics to identify whether there is any non-literal intent;

[0136] Based on the characteristics of the publisher group, the key nodes in the content dissemination path, the diffusion speed, and the non-literal intent, the preliminary classification of the online discussion popularity index signal is revised.

[0137] Based on the revised classification results, preliminary market signals are obtained.

[0138] Specifically, receiving raw market signals from different sources and performing preliminary classification of these raw market signals means that the system obtains raw market information from various channels such as news media, industry reports, social platforms, and sales data, and performs preliminary classification based on preset rules or machine learning models, such as into categories like product promotions, online discussion popularity index, and policy changes.

[0139] When receiving promotional information from competitors, to more accurately understand their true intentions and potential impact, it's necessary to conduct in-depth analysis of the technical specifications differences, product lifecycle stage, and competitors' market strategies. Technical specifications differences refer to comparing the performance, functions, and materials of competitors' promotional products with our own, in order to assess their competitiveness and market positioning. The product lifecycle stage refers to determining whether the promotional product is in the introduction, growth, maturity, or decline phase, which helps predict the sustainability and effectiveness of the promotion. Competitors' market strategies involve analyzing whether their promotional activities are for short-term inventory clearance, long-term market penetration, new product launches, or strategies targeting specific user groups. The aim is to reveal the deeper business logic behind the promotional information, avoiding simple judgments based solely on price.

[0140] Furthermore, based on the aforementioned differences in technical specifications, the product lifecycle stage, and the market strategy, the initial classification of promotional signals is revised. This means that if the initial classification simply categorizes a promotional message as a "price war," but in-depth analysis reveals that it is a strategic promotion for a competitor's new product launch, it should be revised to a more accurate classification such as "new product promotion" or "market share competition" to reflect its true market significance.

[0141] Furthermore, when receiving signals of online discussion popularity on social media, to avoid misinterpreting superficial popularity, it's necessary to analyze the characteristics of the publishers of trending content, key nodes in the content dissemination path, and the speed of diffusion. Publisher characteristics refer to identifying whether the information source is an ordinary user, an industry KOL, a media organization, or an online troll; this helps assess the information's authority and credibility. Key nodes in the content dissemination path identify which accounts or platforms amplify or distort the information during its spread; this helps understand the information's diffusion mechanism. Diffusion speed refers to the breadth and depth of information dissemination within a short period, which helps determine the explosiveness and impact of the online discussion popularity index.

[0142] Simultaneously, deep semantic analysis is performed on the text of the trending content to identify any non-literal intentions. This includes using natural language processing techniques to identify complex semantic phenomena in the text, such as irony, metaphor, and emotional polarity reversal, to distinguish whether seemingly positive or negative statements have the opposite true intention. For example, some seemingly praising statements may actually contain sarcasm, or some seemingly complaining statements may simply be users' expectations for product improvement.

[0143] Therefore, based on the characteristics of the publisher group, the key nodes in the content dissemination path, the diffusion speed, and the non-literal intent, the initial classification of the online discussion heat index signal is revised. For example, if the initial classification categorizes a social media hot topic as "negative product online discussion heat index," but in-depth analysis reveals that it is maliciously hyped by a few competitors' online trolls, and the dissemination path is unnatural, then it should be revised to "malicious competitive interference" or "fake online discussion heat index" to reflect its unrealistic market feedback.

[0144] Ultimately, based on the revised classification results, a preliminary classification of market signals is obtained. This means that after the above detailed analysis and revision, the resulting market signal classification will be more accurate and in-depth, and will be able to more realistically reflect market dynamics and potential risks.

[0145] The proposed solution effectively addresses the superficiality and inaccuracy issues that may exist in the initial classification of market signals in the basic solution by introducing a mechanism for in-depth analysis and correction of specific types of market signals.

[0146] In response, this application further proposes a method for calculating the true impact of market signals based on the original strength of the market signals according to the preliminary classification and multi-dimensional auxiliary scenario information, specifically including:

[0147] Monitor the data update frequency of the multi-dimensional auxiliary scenario information;

[0148] Track the predictive accuracy of the multi-dimensional auxiliary scenario information in historical decision-making events;

[0149] Calculate the real-time effectiveness score of the multi-dimensional auxiliary scenario information based on the data update frequency and the prediction accuracy.

[0150] Based on the real-time effectiveness score, the weight parameters of the multi-dimensional auxiliary scenario information in the scenario weighting function are dynamically adjusted.

[0151] The original strength of the market signals in the preliminary classification and the adjusted weight parameters are applied to the scenario weighting function to calculate the true influence of the market signals.

[0152] Specifically, the data update frequency for monitoring multi-dimensional ancillary scenario information refers to the cycle in which the system continuously monitors and records the release or update of various ancillary scenario information (such as macroeconomic data, industry reports, and consumer behavior trends). The purpose is to ensure that the ancillary scenario information used is up-to-date and to avoid decision-making biases due to data lag. For example, daily updated stock index data has a higher update frequency, while quarterly industry analysis reports have a lower update frequency.

[0153] The tracking of the predictive accuracy of multi-dimensional auxiliary contextual information in historical decision-making events can be understood as the system establishing a historical database to record the degree of consistency between the prediction and the actual outcome after each decision analysis using specific auxiliary contextual information. Its purpose is to evaluate the reliability and effectiveness of different types of auxiliary contextual information. For example, if a certain consumer sentiment index has successfully predicted product sales fluctuations multiple times in the past, its predictive accuracy is considered high.

[0154] In practical applications, calculating the real-time effectiveness score of multi-dimensional auxiliary scenario information based on data update frequency and prediction accuracy means comprehensively evaluating the monitored update frequency and tracked prediction accuracy through a preset algorithm or model to obtain a quantitative score that reflects the effectiveness and reliability of the auxiliary scenario information at the current moment.

[0155] Furthermore, based on the real-time effectiveness score, the weight parameters of multi-dimensional auxiliary scenario information in the scenario weighting function are dynamically adjusted. This means that when the real-time effectiveness score of a certain auxiliary scenario information is high, its weight parameter in the scenario weighting function is increased accordingly, and vice versa. The scenario weighting function is a mathematical model used to calculate the true impact by combining the original strength of market signals and multi-dimensional auxiliary scenario information. By dynamically adjusting the weights, it is ensured that more effective and reliable auxiliary scenario information plays a greater role in the final impact calculation.

[0156] Therefore, applying the original strength of the initially classified market signals and the adjusted weight parameters to the scenario weighting function to calculate the true impact of the market signals means inputting the auxiliary scenario information, after the aforementioned dynamic weight adjustment, along with the original strength of the initially classified market signals, into the scenario weighting function to obtain a more accurate and realistic true impact of the market signals. This scenario weighting function can be a linear or nonlinear combination model, and its specific form can be designed according to actual business needs and data characteristics.

[0157] This application addresses the problem of inconsistent timeliness and accuracy of auxiliary scenario information, leading to biases in the assessment of true impact, inherent in traditional methods, by introducing a real-time effectiveness evaluation mechanism for multi-dimensional auxiliary scenario information. Through this technical solution, the application effectively overcomes the inaccuracy in calculating the true impact of market signals caused by traditional methods failing to fully consider the real-time effectiveness of multi-dimensional auxiliary scenario information. By dynamically evaluating and adjusting the weights of auxiliary scenario information, this application ensures that the scenario information relied upon in the decision analysis process maintains a consistently high level of quality and relevance. This not only significantly improves the accuracy of calculating the true impact of market signals, making decision recommendations more reliable and instructive, but also enhances the adaptability and robustness of the entire big data decision analysis system to changes in the external environment, thereby providing enterprises with more insightful decision support.

[0158] The above-mentioned dynamic adjustment of the weight parameters of multi-dimensional auxiliary scenario information in the scenario weighting function based on real-time effectiveness scores includes the following steps:

[0159] First, the real-time effectiveness score of the multi-dimensional auxiliary scenario information is evaluated. The real-time effectiveness score can be understood as a comprehensive quantitative indicator of the quality, timeliness, accuracy, and contribution of the auxiliary scenario information to the decision-making outcome. The evaluation result directly reflects the reliability of the information in the current decision-making cycle.

[0160] Furthermore, when the real-time effectiveness score falls below a preset negative impact threshold for multiple consecutive monitoring periods, or when the real-time effectiveness score shows a sustained and significant downward trend compared to historical average performance, the multi-dimensional auxiliary scenario information is judged to be inefficient or have a negative impact. The setting of "multiple consecutive monitoring periods" aims to avoid misjudgments caused by short-term data fluctuations and ensure the stability of the judgment. The "preset negative impact threshold" is a critical value pre-set by the system; below this value, it indicates that the auxiliary scenario information may have quality problems or is no longer applicable. "Historical average performance" provides an important benchmark for evaluating the current real-time effectiveness score, while a "sustainable and significant downward trend" clearly indicates that the information quality is systematically deteriorating. The core of this step is to accurately identify auxiliary scenario information that is no longer reliable or may be misleading.

[0161] Once the system determines that the multi-dimensional auxiliary scenario information is inefficient or has a negative impact, it will perform a weight suppression operation. This weight suppression operation includes setting the weight parameter of the multi-dimensional auxiliary scenario information in the scenario weighting function to zero, or multiplying the weight parameter by a preset minimum attenuation factor. Setting the weight parameter to zero means that the auxiliary scenario information will be completely excluded from the scenario weighting function, suitable for situations where the information is completely ineffective or may have a serious negative impact. Multiplying the weight parameter by a preset minimum attenuation factor, such as 0.01 or 0.001, significantly reduces its influence without complete removal, minimizing its potential negative effects. This operation aims to prevent unreliable information from negatively impacting the final decision analysis.

[0162] Simultaneously, the system will record the suppressed multi-dimensional auxiliary scenario information and the reasons for suppression. This recording step is crucial for subsequent system auditing, problem tracing, and the formulation of optimization strategies. The recorded content may include the time when suppression occurred, the specific scenario information identifier, the real-time effectiveness score at that time, the specific conditions that triggered the suppression (e.g., below a threshold or a continuous downward trend), and the type of suppression executed.

[0163] During subsequent monitoring periods, the system will continuously track the real-time effectiveness score of the suppressed multi-dimensional auxiliary scenario information. Even if the information is suppressed, the system will not completely abandon its monitoring, but will continue to pay attention to its performance to determine whether it is possible to restore its effectiveness.

[0164] Finally, when the real-time effectiveness score of the suppressed multi-dimensional auxiliary scenario information returns to the normal range and remains there for a period of time, the system will gradually restore the weight parameters of the multi-dimensional auxiliary scenario information. "Returning to the normal range" means that the score again reaches or exceeds the preset effectiveness standard. "Remaining for a period of time" is to ensure the stability of the recovery and avoid misjudgment due to a temporary rebound. "Gradual recovery" can employ various strategies, such as linear growth, exponential growth, or phased recovery, increasing the weight parameter by a certain percentage each time over several monitoring periods until it reaches its proper weight. This aims to reuse the auxiliary scenario information that has regained its effectiveness, ensuring the comprehensiveness and dynamic adaptability of the decision analysis.

[0165] This application's solution effectively addresses the limitations of the aforementioned basic solutions when processing inefficient or negative auxiliary scenario information by introducing an intelligent weight suppression and recovery mechanism. Through this technical solution, this application can significantly improve the accuracy and robustness of big data decision analysis.

[0166] This application further proposes a step for determining that the multi-dimensional auxiliary scenario information has inefficiency or negative impact when the real-time effectiveness score is lower than a preset negative impact threshold for multiple consecutive monitoring periods, or when the real-time effectiveness score shows a continuous and significant downward trend compared to the historical average performance.

[0167] Obtain the short-term fluctuation range and long-term trend of the real-time effectiveness score;

[0168] Obtain a real-time effectiveness score for scenario information of the same type as the multi-dimensional auxiliary scenario information;

[0169] When the short-term fluctuation exceeds the preset short-term fluctuation threshold, the decrease in the real-time effectiveness score is determined to be short-term market noise.

[0170] When the long-term trend indicates that the real-time effectiveness score is continuously declining, the decline in the real-time effectiveness score is determined to be a long-term trend.

[0171] When the real-time effectiveness score is lower than the average score of the same type of scenario information, and the average score of the same type of scenario information remains stable, it is determined that the multi-dimensional auxiliary scenario information is inefficient or has a negative impact.

[0172] Based on the judgment of short-term market noise, the judgment of long-term trends, and the judgment of the average score of similar scenario information, it is determined whether the multi-dimensional auxiliary scenario information has inefficiency or negative impact.

[0173] Specifically, obtaining the short-term fluctuation range and long-term trend of the real-time validity score refers to quantifying the range of variation of the real-time validity score within a short time window and its overall trend within a longer time window through statistical analysis methods, such as moving averages and standard deviation calculations. The short-term fluctuation range can be understood as the difference between the maximum and minimum values ​​of the score within a specific time period, or its standard deviation, aiming to identify instantaneous disturbances in the score. The long-term trend can be fitted using methods such as linear regression and exponential smoothing to reveal the continuous rise, fall, or stable state of the score, aiming to distinguish between random fluctuations and structural changes.

[0174] Furthermore, obtaining real-time validity scores for scenario information of the same type as the multi-dimensional auxiliary scenario information means that the system maintains a scenario information type library and independently monitors the real-time validity scores for each type of scenario information. When it is necessary to judge specific multi-dimensional auxiliary scenario information, the system will search for its type and obtain the real-time validity scores of all other scenario information under that type. The purpose is to provide an objective reference benchmark and avoid misjudgments due to the particularity of a single scenario information.

[0175] Specifically, when the short-term fluctuation exceeds a preset short-term fluctuation threshold, the decline in the real-time effectiveness score is considered short-term market noise. This means that if the score decline is only a brief and sharp fluctuation without forming a sustained trend, it is regarded as a temporary disturbance in the market or data collection process, rather than a fundamental problem with the effectiveness of the contextual information itself. This short-term fluctuation threshold can be set based on historical data and business experience. When the long-term trend indicates a continuous decline in the real-time effectiveness score, the decline is considered a long-term trend, indicating that the decline in the score is stable and continuous. This usually suggests that the effectiveness of the multi-dimensional auxiliary contextual information may indeed be weakening, requiring attention.

[0176] In practical applications, when the real-time effectiveness score is lower than the average score of similar scenario information, and the average score of similar scenario information remains stable, the multi-dimensional auxiliary scenario information is judged to be inefficient or have a negative impact. This provides a comparative perspective; if the performance of a certain scenario information is significantly lower than its peer average, and the peer average itself has not decreased, it indicates that the specific scenario information may have its own problems, rather than a general problem of the entire market environment or data source. Ultimately, based on the judgment of short-term market noise, the judgment of long-term trends, and the judgment of the average score of similar scenario information, a comprehensive determination is made as to whether the multi-dimensional auxiliary scenario information is inefficient or has a negative impact. This means that the system will combine the analysis results of these three dimensions to make a multi-factor comprehensive decision to improve the accuracy and robustness of the judgment. For example, only when short-term noise is excluded, the long-term trend does indeed decline, and it is lower than the peer average, is it ultimately judged to be inefficient or have a negative impact.

[0177] This application's solution addresses the potential misjudgment problem in the basic solution by introducing multi-dimensional judgment criteria. Through the above technical solution, this application can significantly improve the accuracy and robustness of judging the validity of multi-dimensional auxiliary contextual information.

[0178] This application further proposes the steps for determining whether multi-dimensional auxiliary scenario information is inefficient or has a negative impact when the real-time effectiveness score is lower than a preset negative impact threshold for multiple consecutive monitoring periods, or when the real-time effectiveness score shows a continuous and significant downward trend compared to the historical average performance. Specifically, these steps include:

[0179] Obtain the type identifier of the multi-dimensional auxiliary scenario information;

[0180] Based on the type identifier, load the negative impact threshold and downward trend judgment time window corresponding to the type identifier from the preset type rule base;

[0181] The real-time effectiveness score is compared with the negative impact threshold of the loading.

[0182] Within the loaded downward trend judgment time window, analyze the changes in the real-time effectiveness score to determine whether the real-time effectiveness score has a continuous downward trend;

[0183] Based on the comparison and analysis results, it is determined whether the multi-dimensional auxiliary scenario information has inefficiency or negative impact.

[0184] Specifically, the type identifier for acquiring multi-dimensional auxiliary scenario information refers to the category to which the scenario information currently being evaluated belongs, either automatically identified by the system or manually specified, such as "macroeconomic data," "industry competitive intelligence," or "consumer sentiment index." This type identifier forms the basis for subsequently loading specific judgment rules. Loading the corresponding negative impact threshold and downward trend judgment time window from a pre-set type rule library based on the type identifier can be understood as the system maintaining a database containing multiple scenario information types, each with its own pre-configured judgment parameters.

[0185] In practical applications, comparing the real-time effectiveness score with the negative impact threshold means that the system directly compares the currently calculated real-time effectiveness score with the negative impact threshold specific to this type of information to determine whether it has reached the preset level of inefficiency or negativity. Furthermore, within the downward trend judgment time window, the system analyzes the changes in the real-time effectiveness score to determine whether there is a continuous downward trend. The purpose is to identify the long-term trend of the score rather than short-term fluctuations. For example, for a downward trend judgment time window set at "five consecutive monitoring periods," the system will check whether the score shows a continuous and significant decline within these five periods, rather than just an occasional decline within a single period.

[0186] Therefore, determining whether multi-dimensional auxiliary scenario information is inefficient or has a negative impact based on comparison and analysis results means comprehensively considering whether the score is below the threshold and whether there is a continuous downward trend, and finally making a judgment on the effectiveness of the scenario information.

[0187] This application's solution, by introducing type identifiers and a type rule base, moves away from using a single, universal standard to judge the effectiveness of multi-dimensional auxiliary contextual information. Instead, it dynamically loads specific negative impact thresholds and downward trend judgment time windows based on the inherent characteristics of different types of information. This addresses the limitations inherent in the effectiveness of multi-dimensional auxiliary contextual information.

[0188] When the real-time effectiveness score is below the preset negative impact threshold for multiple consecutive monitoring periods, or when the real-time effectiveness score shows a continuous and significant downward trend compared to the historical average performance, the multi-dimensional auxiliary scenario information is judged to be inefficient or have a negative impact, including:

[0189] Obtain the average level and fluctuation range of the multi-dimensional auxiliary scenario information over a period of time before the event occurs;

[0190] Retrieve the type, scope of impact, and estimated duration of currently occurring uncontrollable events;

[0191] Based on the type of the uncontrollable event, load the impact pattern of the type on different scenario information;

[0192] Compare the actual performance of the multi-dimensional auxiliary contextual information during the event with the impact pattern of its loading;

[0193] Determine whether the magnitude and duration of the score decline in the multi-dimensional auxiliary scenario information are consistent with the short-term impact characteristics predicted by the impact pattern, and whether the score decline exceeds the average level and fluctuation range before the event occurred;

[0194] Based on the assessment results, the long-term strategic value of the multi-dimensional auxiliary scenario information is re-evaluated.

[0195] Specifically, before determining whether multi-dimensional auxiliary scenario information is inefficient or has a negative impact, it is first necessary to obtain the average level and fluctuation range of this multi-dimensional auxiliary scenario information over a period of time before the occurrence of the current uncontrollable event. This average level and fluctuation range can be understood as the baseline performance of the scenario information under normal market conditions, and are used for subsequent comparative analysis. For example, the mean and standard deviation of the real-time effectiveness score of the scenario information over the past few weeks or months can be calculated.

[0196] Simultaneously, the system will retrieve the type, scope of impact, and estimated duration of currently occurring uncontrollable events. Uncontrollable events may include, but are not limited to, adjustments to macroeconomic policies, changes in industry regulations, major natural disasters, public health emergencies, large-scale social network discussion crises, or disruptive innovations by major competitors. This information can be obtained in real time through external data sources (such as news media, government announcements, and industry reports).

[0197] Furthermore, based on the type of uncontrollable event retrieved, the system loads the potential impact patterns of that type of event on different contextual information from a pre-set knowledge base or model library. The impact pattern can be a set of parameters describing the typical magnitude, duration, and recovery path of the potential decline in contextual information scores caused by a specific type of event. For example, a policy adjustment might cause a 20% drop in contextual information scores for a specific industry in the short term, lasting for three months.

[0198] The system then compares the actual performance of the multi-dimensional auxiliary scenario information during uncontrollable events (i.e., changes in real-time effectiveness scores) with the loaded impact patterns. This comparison aims to assess whether the observed score changes align with the expected impact of this type of event.

[0199] Based on this, the system will determine whether the magnitude and duration of the score decline in the multi-dimensional auxiliary scenario information are consistent with the short-term impact characteristics predicted by the impact pattern, and whether the score decline exceeds the average level and fluctuation range before the event. If the characteristics of the score decline are consistent with the short-term impact predicted by the impact pattern, and do not significantly exceed the normal fluctuation range before the event, it indicates that the decline may be mainly caused by uncontrollable events, rather than a problem with the long-term value of the scenario information itself.

[0200] Ultimately, based on the above assessment, the system will re-evaluate the long-term strategic value of multi-dimensional auxiliary scenario information. This means that even if the score temporarily declines, if it is judged as a short-term shock, the scenario information may still be considered to have long-term value, thus avoiding being wrongly suppressed.

[0201] This application's solution optimizes the accuracy of multi-dimensional auxiliary scenario information validity judgment by introducing a mechanism for identifying and assessing the impact of uncontrollable events. Through this technical solution, the application can effectively distinguish between short-term fluctuations caused by uncontrollable events and the long-term inefficiency or negative impact of the scenario information itself. This avoids unnecessary weight suppression of scenario information with long-term value, thus ensuring that the decision analysis system can continuously utilize all relevant and effective data sources. Furthermore, by reassessing the long-term strategic value of scenario information, this solution helps maintain the stability and adaptability of the decision model, enabling it to make more accurate and comprehensive decisions in the face of complex and ever-changing market environments, significantly improving the reliability and practicality of big data decision analysis methods.

[0202] refer to Figure 3 , Figure 3 This is a schematic diagram of the structure of a big data decision analysis system provided in an embodiment of the present invention, including:

[0203] The receiving end is used to receive market signals from different sources and classify the market signals to obtain preliminarily classified market signals.

[0204] The computing end is used to initiate the process of collecting auxiliary scenario information for the market signals that are initially classified; to acquire multi-dimensional auxiliary scenario information related to the market signals that are initially classified; and to calculate the real influence of the market signals based on the original strength of the market signals that are initially classified and the multi-dimensional auxiliary scenario information.

[0205] The identification end is used to sort conflicting market signals based on their actual influence and identify the dominant market signals.

[0206] The suggestion side is used to generate decision-making suggestions based on the dominant market signals and their actual influence.

[0207] This system employs a modular design, implementing core functions such as market signal reception, contextual information collection and influence calculation, contradictory signal identification and ranking, and decision recommendation generation through independent components. This creates an efficient, accurate, and robust big data decision analysis framework. By introducing multi-dimensional auxiliary contextual information and a real influence calculation mechanism, the system can effectively identify and process contradictory market signals, thereby generating more stable and accurate decision recommendations and avoiding decision lags and errors caused by improper information processing in traditional methods.

[0208] The specific steps and working principles of the big data decision analysis method have been described in the above embodiments and will not be repeated here. It should be emphasized that the big data decision analysis system proposed in this application, through its specific component configuration, concretizes the above method steps into operable system modules.

[0209] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A big data decision analysis method, characterized in that, include: Receive market signals from different sources and classify the market signals to obtain preliminarily classified market signals; For the market signals identified in the preliminary classification, the process of collecting auxiliary scenario information is initiated; Obtain multi-dimensional auxiliary scenario information related to the market signals of the preliminary classification; Based on the original strength of the market signals according to the preliminary classification and the multi-dimensional auxiliary contextual information, the true influence of the market signals is calculated. Text content on social media platforms is received, and preliminary sentiment identification is performed on the text content to obtain a preliminary sentiment tendency value. The historical content style profile of the publisher of the text content is queried. Based on the frequency of use of rhetorical questions, exaggerated words, and puns in the historical content style profile, as well as the proportion of specific emojis appearing in the comment section interaction triggered by the publisher's content, the publisher's irony tendency index is calculated. Comment section interaction data related to the text content is monitored and collected. Specific emojis that appear frequently in the comment section interaction data are statistically analyzed, and keywords and phrases containing questioning, rhetorical questions, puns, or those that contradict the surface meaning of the text content are identified to calculate the interaction sentiment deviation index. Within a preset period after the text content is published, it is continuously monitored whether the text content is used for secondary creation, meme dissemination, or over-interpretation by other users to calculate the deconstruction dissemination index. The preliminary sentiment tendency value, the publisher's irony tendency index, the interaction sentiment deviation index, and the deconstruction dissemination index are used as inputs, and the calibrated sentiment tendency is calculated through a sentiment polarity correction function. Based on the calibrated sentiment tendency, the influence weight of social media signals in the overall decision analysis is adjusted, and the true influence of the market signals is calculated. Based on the stated true influence, conflicting market signals are ranked, and the dominant market signals are identified. Based on the dominant market signals and their actual influence, decision-making recommendations are generated.

2. The big data decision analysis method according to claim 1, characterized in that, The generation of decision recommendations based on the dominant market signals and their actual influence includes: Assess the potential sales benefits, potential risks of unsold inventory, and costs associated with increasing inventory levels. Assess the potential cost savings, potential sales losses, and market share loss risks associated with reducing the aforementioned inventory decision direction; By comparing the potential sales revenue, the potential unsold inventory risk and cost, the potential cost savings, the potential sales loss, and the market share loss risk, and combining these with preset business objectives, a risk-adjusted decision score is generated. Based on the risk-adjusted decision score, decision recommendations are generated.

3. The big data decision analysis method according to claim 1, characterized in that, The process of receiving market signals from different sources and classifying the market signals to obtain preliminarily classified market signals includes: Receive raw market signals from different sources and perform preliminary classification of the raw market signals; When receiving promotional information from competitors, analyze the differences in technical specifications of the promoted products, the product life cycle stage, and the competitors' market strategies. Based on the aforementioned differences in technical specifications, the aforementioned product life cycle stage, and the aforementioned market strategy, the initial classification of promotional signals is revised; When receiving a trending index signal on social media, analyze the characteristics of the group of publishers of trending content, key nodes in the content dissemination path, and the speed of diffusion; Perform deep semantic analysis on the text of the hot topics to identify whether there is any non-literal intent; Based on the characteristics of the publisher group, the key nodes in the content dissemination path, the diffusion speed, and the non-literal intent, the preliminary classification of the online discussion popularity index signal is revised. Based on the revised classification results, preliminary market signals are obtained.

4. The big data decision analysis method according to claim 1, characterized in that, The calculation of the true impact of the market signal based on the original strength of the market signal according to the preliminary classification and the multi-dimensional auxiliary scenario information includes: Monitor the data update frequency of the multi-dimensional auxiliary scenario information; Track the predictive accuracy of the multi-dimensional auxiliary scenario information in historical decision-making events; Calculate the real-time effectiveness score of the multi-dimensional auxiliary scenario information based on the data update frequency and the prediction accuracy. Based on the real-time effectiveness score, the weight parameters of the multi-dimensional auxiliary scenario information in the scenario weighting function are dynamically adjusted. The original strength of the market signals in the preliminary classification and the adjusted weight parameters are applied to the scenario weighting function to calculate the true influence of the market signals.

5. The big data decision analysis method according to claim 4, characterized in that, The step of dynamically adjusting the weight parameters of the multi-dimensional auxiliary scenario information in the scenario weighting function based on the real-time effectiveness score includes: Evaluate the real-time effectiveness score of the multi-dimensional auxiliary scenario information; When the real-time effectiveness score is lower than the preset negative impact threshold for multiple consecutive monitoring periods, or when the real-time effectiveness score shows a continuous and significant downward trend compared with the historical average performance, the multi-dimensional auxiliary scenario information is judged to be inefficient or have a negative impact. Perform a weight suppression operation, which includes setting the weight parameter of the multi-dimensional auxiliary scenario information in the scenario weighting function to zero, or multiplying the weight parameter by a preset minimum attenuation factor; Record the suppressed multi-dimensional auxiliary scenario information and the reasons for its suppression; During subsequent monitoring periods, the real-time effectiveness score of the suppressed multi-dimensional auxiliary scenario information will be continuously tracked. When the real-time effectiveness score of the suppressed multi-dimensional auxiliary scenario information returns to the normal range and remains so for a period of time, the weight parameters of the multi-dimensional auxiliary scenario information are gradually restored.

6. The big data decision analysis method according to claim 5, characterized in that, When the real-time effectiveness score is below a preset negative impact threshold for multiple consecutive monitoring periods, or when the real-time effectiveness score shows a continuous and significant downward trend compared to the historical average performance, the multi-dimensional auxiliary scenario information is determined to be inefficient or have a negative impact, including: Obtain the short-term fluctuation range and long-term trend of the real-time effectiveness score; Obtain a real-time effectiveness score for scenario information of the same type as the multi-dimensional auxiliary scenario information; When the short-term fluctuation exceeds the preset short-term fluctuation threshold, the decrease in the real-time effectiveness score is determined to be short-term market noise. When the long-term trend indicates that the real-time effectiveness score is continuously declining, the decline in the real-time effectiveness score is determined to be a long-term trend. When the real-time effectiveness score is lower than the average score of the same type of scenario information, and the average score of the same type of scenario information remains stable, it is determined that the multi-dimensional auxiliary scenario information is inefficient or has a negative impact. Based on the judgment of short-term market noise, the judgment of long-term trends, and the judgment of the average score of similar scenario information, it is determined whether the multi-dimensional auxiliary scenario information has inefficiency or negative impact.

7. The big data decision analysis method according to claim 5, characterized in that, When the real-time effectiveness score is below a preset negative impact threshold for multiple consecutive monitoring periods, or when the real-time effectiveness score shows a continuous and significant downward trend compared to the historical average performance, the multi-dimensional auxiliary scenario information is judged to be inefficient or have a negative impact, including: Obtain the type identifier of the multi-dimensional auxiliary scenario information; Based on the type identifier, load the negative impact threshold and downward trend judgment time window corresponding to the type identifier from the preset type rule base; The real-time effectiveness score is compared with the negative impact threshold of the loading. Within the loaded downward trend judgment time window, analyze the changes in the real-time effectiveness score to determine whether the real-time effectiveness score has a continuous downward trend; Based on the comparison and analysis results, it is determined whether the multi-dimensional auxiliary scenario information has inefficiency or negative impact.

8. The big data decision analysis method according to claim 5, characterized in that, When the real-time effectiveness score is below a preset negative impact threshold for multiple consecutive monitoring periods, or when the real-time effectiveness score shows a continuous and significant downward trend compared to the historical average performance, the multi-dimensional auxiliary scenario information is determined to be inefficient or have a negative impact, including: Obtain the average level and fluctuation range of the multi-dimensional auxiliary scenario information over a period of time before the event occurs; Retrieve the type, scope of impact, and estimated duration of currently occurring uncontrollable events; Based on the type of the uncontrollable event, load the impact pattern of the type of the uncontrollable event on different scenario information; Compare the actual performance of the multi-dimensional auxiliary contextual information during the event with the impact pattern of its loading; Determine whether the magnitude and duration of the score decline in the multi-dimensional auxiliary scenario information are consistent with the short-term impact characteristics predicted by the loaded impact pattern, and whether the score decline exceeds the average level and fluctuation range before the event occurred; Based on the assessment results, the long-term strategic value of the multi-dimensional auxiliary scenario information is re-evaluated.

9. A big data decision analysis system, characterized in that, include: The receiving end is used to receive market signals from different sources and classify the market signals to obtain preliminarily classified market signals. The computing end is used to initiate the process of collecting auxiliary scenario information based on the market signals of the preliminary classification; Obtain multi-dimensional auxiliary scenario information related to the market signals of the preliminary classification; Based on the original strength of the market signals according to the preliminary classification and the multi-dimensional auxiliary contextual information, the true influence of the market signals is calculated. Text content on social media platforms is received, and preliminary sentiment identification is performed on the text content to obtain a preliminary sentiment tendency value. The historical content style profile of the publisher of the text content is queried. Based on the frequency of use of rhetorical questions, exaggerated words, and puns in the historical content style profile, as well as the proportion of specific emojis appearing in the comment section interaction triggered by the publisher's content, the publisher's irony tendency index is calculated. Comment section interaction data related to the text content is monitored and collected. Specific emojis that appear frequently in the comment section interaction data are statistically analyzed, and keywords and phrases containing questioning, rhetorical questions, puns, or those that contradict the surface meaning of the text content are identified to calculate the interaction sentiment deviation index. Within a preset period after the text content is published, it is continuously monitored whether the text content is used for secondary creation, meme dissemination, or over-interpretation by other users to calculate the deconstruction dissemination index. The preliminary sentiment tendency value, the publisher's irony tendency index, the interaction sentiment deviation index, and the deconstruction dissemination index are used as inputs, and the calibrated sentiment tendency is calculated through a sentiment polarity correction function. Based on the calibrated sentiment tendency, the influence weight of social media signals in the overall decision analysis is adjusted, and the true influence of the market signals is calculated. The identification end is used to sort conflicting market signals based on their actual influence and identify the dominant market signals. The suggestion side is used to generate decision-making suggestions based on the dominant market signals and their actual influence.

Images