System and method for market analysis and automated business decision

Abstract

The invention relates to the field of market management, provides a market analysis and automatic business decision system and method, and aims to improve the dynamic response speed and decision quality of an enterprise to a market, and the method comprises the steps: building a prediction model based on historical data related to target market analysis or business decision and a corresponding bipartite type target; determining a threshold value for business decision through a model performance evaluation method; and determining a decision rule in combination with a threshold. The data response speed is increased by establishing a prediction model; and the decision-making quality is improved by determining the dynamic threshold value.

Description

Technical Field

[0001] The present invention relates to the field of market management, and specifically to a system and method for market analysis and automated business decision-making. Background Art

[0002] With the increasingly fierce market competition and the rapid changes in the business environment, enterprises have put forward higher requirements for the efficiency and accuracy of market analysis and business decision-making. Market analysis involves the collection and analysis of information on various aspects such as consumer needs, market trends, and competitors, while business decision-making requires formulating corresponding strategies based on these analysis results.

[0003] Traditional market analysis methods usually collect market data through means such as questionnaires, social media monitoring, and sales data recording. However, these methods are often time-consuming and laborious, and it is difficult to guarantee the quality and integrity of the data; in addition, with the rapid growth of data volume, traditional data collection methods can no longer meet the needs of enterprises for big data processing. In recent years, although some data analysis tools and technologies have emerged, there are still deficiencies in aspects such as data integration, analysis depth, and real-time performance. In terms of decision support, existing decision support systems are usually built based on expert systems or rule engines, and assist in decision-making through preset rules and logics; however, these systems often lack flexibility and adaptability and cannot cope with the changing market environment and complex business scenarios. Summary of the Invention

[0004] In order to improve the response speed of enterprises to market dynamics and the quality of decision-making, the present application provides a system and method for market analysis and automated business decision-making.

[0005] The technical solution adopted by the present invention to solve the above problems is as follows:

[0006] A method for market analysis and automated business decision-making, including:

[0007] Step 1: Obtain historical data related to target market analysis or business decision-making and its corresponding binary type targets; the historical data includes sample objects and their corresponding attributes, and the sample objects are products, consumers, or market regions; the attributes are sales volume, price, consumer satisfaction, and market share;

[0008] Step 2: Based on a machine learning algorithm, establish a prediction model with the sample objects and their corresponding attributes as inputs and the binary type targets as outputs, and train it;

[0009] Step 3: Determine the threshold for business decision-making based on a model performance evaluation method;

[0010] Step 4: Input new data into the prediction model for prediction, and determine the decision rule in combination with the threshold.

[0011] Further, step 1 also includes preprocessing the collected historical data.

[0012] Further, the preprocessing includes: removing duplicate data, filling in missing values, correcting outliers, and unifying the data format.

[0013] Further, the model performance evaluation method is the pb value or the F1 score.

[0014] Further, step 4 also includes: determining whether the input new data is emergency data; if it is emergency data, then giving priority to predicting the emergency data.

[0015] Further, step 1 also includes dividing the historical data into emergency data and non-emergency data, creating a data type recognition model based on the division result; step 4 determines whether the input new data is emergency data based on the data type recognition model.

[0016] Further, it also includes step 5: regularly optimizing the prediction model.

[0017] Further, the optimization methods include:

[0018] A. Monitoring the decision execution situation and optimizing the prediction model according to the monitoring results; and / or

[0019] B. Introducing new data to retrain the prediction model.

[0020] A system for market analysis and automated business decision-making, including:

[0021] A data collection module: obtaining historical data related to target market analysis or business decision-making and its corresponding binary type target;

[0022] A prediction model establishment module: based on machine learning algorithms, using sample objects and their corresponding attributes as inputs and binary type targets as outputs to establish and train a prediction model;

[0023] A threshold determination module: determining the threshold for business decision-making based on the model performance evaluation method;

[0024] A prediction module: obtaining new data, inputting the new data into the prediction model for prediction, and combining the threshold to determine the decision rule.

[0025] Further, it also includes: a data type recognition module: used to determine whether new data is emergency data; when the new data is determined to be emergency data, giving priority to predicting the emergency data.

[0026] The beneficial effects of the present invention compared with the prior art are as follows: a prediction model is established based on historical data related to target market analysis or business decision-making and its corresponding binary type targets; a threshold for business decision-making is determined through a model performance evaluation method; and a decision rule is determined in combination with the threshold. The data response speed is improved by establishing the prediction model; and the decision-making quality is improved by determining the dynamic threshold. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flowchart of a method for market analysis and automated business decision-making. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0029] As Figure 1 shown, a method for market analysis and automated business decision-making includes:

[0030] Step 1: Obtain historical data related to target market analysis or business decision-making and its corresponding binary type targets; the historical data includes sample objects and their corresponding attributes, and the sample objects are products, consumers, market regions, etc.; the attributes are sales volume, price, consumer satisfaction, market share, etc.

[0031] Among them, the sales volume data can help understand the market performance of products, the price data can reveal the market competition situation, and the consumer satisfaction data can insight into the needs and preferences of consumers. The corresponding binary type targets refer to whether the sales volume meets the standard, whether the price is appropriate, whether the consumers are satisfied, whether the market share is reasonable, etc. By setting the binary type targets, it can be directly known whether the decision is appropriate.

[0032] To improve the efficiency and quality of data collection, this solution adopts multi-channel data integration technology to connect to the enterprise's internal business systems, third-party data providers, social media platforms, etc., and automatically collects data, covering multi-dimensional information such as sales records, inventory levels, market research feedback, social media hot topics, industry reports, etc., to avoid the inefficiency and errors of manual data collection. Cloud computing and distributed architecture are used to process large amounts of data, and large amounts of data are quickly stored and processed to meet the enterprise's requirements for big data processing.

[0033] However, the original data often has various problems, such as missing values, outliers, inconsistent data formats, etc. Therefore, it is necessary to clean, organize and preprocess the data. This includes steps such as removing duplicate data, filling in missing values, correcting outliers, and unifying data formats. In addition, it is necessary to ensure that the data structure type is consistent with the data structure type in the target table for subsequent data analysis and model training. When processing the original data, machine learning algorithms are introduced to assist in data cleaning, automatically identify and correct outliers, fill in missing values, identify noise and outliers in the data through intelligent algorithms to ensure the accuracy and reliability of the data, and at the same time use data lineage analysis technology to trace the data source and change history to ensure data consistency and traceability.

[0034] Step 2: Based on machine learning algorithms, use the sample object and its corresponding attributes as input, and the binary classification target as output to establish and train a prediction model.

[0035] Machine learning algorithms such as logistic regression, decision tree, random forest, etc. These algorithms can be trained with the sample object and its corresponding attributes as input and the binary classification target T as output, so as to obtain a prediction model that can predict the binary classification target of each object in the target table. By processing the attribute data corresponding to the sample object, the probability value P of the sample object relative to its corresponding binary classification target T is obtained, and the probability value P ∈ (0, 1). The numerical interval from 0 to 1 is equally divided into a preset number of parts x, and each equal division value is assigned to T1 to T in the order from 0 to 1 x 。

[0036] Step 3: Determine the threshold for business decision-making based on the model performance evaluation method.

[0037] After obtaining the prediction model, it is necessary to evaluate its performance and select the best threshold for subsequent business decisions. There are many methods to evaluate the model performance, such as accuracy, recall rate, etc. However, these methods often can only reflect the performance of the model under a certain fixed threshold. In order to find the best threshold, more refined evaluation methods need to be used, such as pb value calculation, F1 score, etc.

[0038] The pb value is an evaluation method that comprehensively considers the model prediction probability and the actual business loss. It can calculate the profit or loss of the model under different situations according to different thresholds and find the best threshold that maximizes the profit or minimizes the loss. Selecting this best threshold as the boundary value can more accurately judge whether the prediction result meets the business requirements, so as to formulate more reasonable business decision-making rules.

[0039] For example, classify and label the sample objects as TP, FP, FN, TN:

[0040] TP: Successfully predict market growth or consumer purchase behavior;

[0041] FP: Error in predicting market growth or consumer purchasing behavior;

[0042] FN: Market growth or consumer purchasing behavior that was not predicted;

[0043] TN: Correct prediction of no market growth or no consumer purchase.

[0044] Statistically obtain the number of sample objects N corresponding to TP TP , the number of sample objects N corresponding to FP FP , the number of sample objects N corresponding to FN FN , the number of sample objects N corresponding to TN TN .

[0045] a = N TP / N (proportion of successfully predicted market growth);

[0046] b = N FP / N (proportion of wrongly predicted market growth);

[0047] c = N FN / N (proportion of market growth that was not predicted);

[0048] d = N TN / N (proportion of correctly predicted no market growth);

[0049] The calculation of the benefit value (pb) can be adjusted based on the specific objectives of the market analysis. For example, it can be an objective function based on prediction accuracy, revenue maximization, or cost minimization. For example,

[0050]

[0051] Where:

[0052] Revenue: Revenue brought about by successfully predicting market growth or consumer purchasing behavior;

[0053] Savings: Cost savings from correctly predicting no market growth or no consumer purchase;

[0054] Total cost: Includes the cost of taking action and the cost of wrong prediction;

[0055] Opportunity cost: Cost resulting from market opportunities that were not predicted.

[0056] For the calculation of the cut-off value: Select the T corresponding to the maximum benefit value (pb) x as the cut-off value T 分 .

[0057] The F1 score is the harmonic mean of precision and recall, which comprehensively considers the prediction accuracy of the model for positive and negative classes. Therefore, in some cases, it can replace the pb value.

[0058] For each threshold T x , use the prediction model to predict each object in the sample dataset to obtain probability values; according to the probability values and the threshold, divide the sample objects into positive classes (predicted value greater than or equal to T x ) and negative classes (predicted value less than T x ); count the quantities of TP, FP, FN, and TN; calculate precision and recall:

[0059] Precision = TP / (TP + FP)

[0060] Recall = TP / (TP + FN)

[0061] Calculate the F1 score:

[0062] F1 score = 2 * (Precision * Recall) / (Precision + Recall)

[0063] Select the threshold corresponding to the maximum F1 score as the cut-off value for subsequent business decisions.

[0064] Step 4: Input the new data into the prediction model for prediction and determine the decision rules in combination with the threshold. For example, when the probability of predicting that the sales volume does not meet the standard exceeds a certain threshold, the system can automatically adjust the price or promotion strategy to stimulate sales growth; when the trend of the decline in customer satisfaction is obvious, the system can trigger a market research or product improvement plan to improve the customer experience. These decision rules can be flexibly adjusted according to the actual needs of the enterprise and the market competition situation to determine the final decision rules.

[0065] A liquor company wants to decide whether to launch a new high-end liquor during the upcoming Spring Festival, and they need to evaluate the potential benefits and risks of this decision.

[0066] Assume that when the advertising investment is 1 million yuan, the model predicts:

[0067] a = 0.7 (70% of the sales successes are correctly predicted);

[0068] b = 0.1 (10% of the sales failures are wrongly predicted as successes);

[0069] c = 0.15 (15% of the sales successes are not predicted);

[0070] d = 0.05 (5% of the sales failures are correctly predicted).

[0071] Calculation of the benefit value (pb):

[0072] Assume the following costs and benefits:

[0073] Benefits: The average benefit from a successful sale is 500,000 yuan;

[0074] Savings: The average cost saved by correctly predicting no sale is 200,000 yuan;

[0075] Total cost: Advertising investment and other related costs are 1,000,000 yuan;

[0076] Opportunity cost: The opportunity cost of missed sales opportunities is 300,000 yuan.

[0077] Use the benefit value formula:

[0078]

[0079]

[0080] pb ≈ 0.419

[0081] Determination of the break-even value:

[0082] By analyzing the pb values at different advertising investment levels, determine the optimal advertising investment threshold that maximizes the benefit. For example, if increasing the advertising investment leads to a decrease in the pb value, then the current investment of 1,000,000 yuan may be the best choice. Based on the calculated benefit value pb, the wine company can decide whether to launch a new high-end white wine during the Spring Festival and determine the optimal advertising investment level. If the pb value is higher than a preset threshold (e.g., 0.4), then launching the new product is considered a profitable decision.

[0083] In addition, when inputting new data into the target model, the system will give priority to processing urgent data according to the data urgency. Due to the large volume and complexity of market data, it takes more time and resources during the processing. By presetting some data as urgent data in advance, urgent data usually requires quick response and processing because they may be directly related to the company's core business, customer satisfaction, market opportunities or potential threats (such as market fluctuations, supply chain disruptions, customer churn). Through classification, the company can ensure that limited resources are preferentially allocated to processing urgent data, thereby improving the overall operational efficiency; and timely identifying and processing these urgent data helps the company take measures in advance to mitigate the impact of risks and even avoid the occurrence of risks.

[0084] Urgent data usually contains the latest information on market changes, customer needs or competitor behavior. These data are crucial for making timely and effective business decisions. Through classification, the company can ensure that the decision-making process is based on the most accurate and relevant data, thereby improving the quality of decisions.

[0085] Specifically, emergency data can be realized by establishing a data type recognition model. First, it is necessary to determine which business metrics fall within the scope of emergency data. For example, when there are significant fluctuations or anomalies in sales, profit margins, customer satisfaction, etc., immediate analysis and response are required.

[0086] Through in-depth analysis of historical emergency data, key features associated with emergency data are extracted. These features are used as training data to train the data type recognition model. The data type recognition model can be established based on machine learning algorithms such as logistic regression, support vector machines, and random forests. The probability prediction value of the occurrence of emergency data can be obtained through the trained model, and this prediction value will be an important basis for judging whether emergency data will appear. To ensure the accuracy of the prediction, a reasonable prediction threshold needs to be set for the emergency data prediction model. When the prediction probability exceeds this threshold, it is judged that emergency data will appear. At this time, through the pre-constructed efficient feedback mechanism (such as building a fast upload channel specifically for emergency data), the emergency data is quickly uploaded, thus realizing the priority processing of emergency data.

[0087] The market environment and business requirements are constantly changing. Therefore, it is necessary to continuously optimize and iterate the relevant models to maintain their effectiveness and adaptability. This includes collecting actual data after decision execution for evaluating decision-making effects, feeding the actual data back into the system for continuous optimization and iteration of the model, regularly evaluating model performance and updating the model in a timely manner, and introducing new data sources and features to improve the prediction ability and accuracy of the model, etc.

[0088] During the process of model optimization and iteration, it is necessary to pay attention to the changes in model performance and business requirements. If the model performance deteriorates or the business requirements change, it is necessary to adjust the model parameters or retrain the model in a timely manner to adapt to the new market environment. At the same time, it is also necessary to continuously explore new data sources and features to enrich the input information of the model and improve the prediction ability of the model. In addition, it is necessary to adjust decision-making rules and business processes according to market changes and business requirements to ensure the effectiveness and adaptability of decision-making. This includes adjusting the triggering conditions of decision-making rules, optimizing business processes to improve decision execution efficiency, etc. Through these adjustments and optimizations, the final decision can better meet the actual needs of the enterprise and the market environment, thus achieving better business results.

[0089] Correspondingly, this embodiment also provides a system for market analysis and automated business decision-making, including:

[0090] Data collection module: Obtain historical data related to target market analysis or business decision-making and their corresponding binary type targets;

[0091] Prediction model building module: Based on machine learning algorithms, a prediction model is built and trained with sample objects and their corresponding attributes as input and binary type targets as output;

[0092] Threshold determination module: Determine the threshold for business decisions based on model performance evaluation methods;

[0093] Prediction module: Obtain new data, input the new data into the prediction model for prediction, and determine the decision rule in combination with the threshold.

[0094] Furthermore, it also includes: Data type identification module: Used to determine whether the new data is emergency data; when the new data is determined to be emergency data, the emergency data is preferentially predicted.

Claims

1. A method for market analysis and automated business decision-making, characterized in that, Including: Step 1: Obtain historical data related to target market analysis or business decision-making and their corresponding binary-type targets; the historical data includes sample objects and their corresponding attributes, where the sample objects are products, consumers, or market regions; the attributes are sales volume, price, consumer satisfaction, and market share; Step 2: Based on machine learning algorithms, use the sample objects and their corresponding attributes as inputs and the binary-type targets as outputs to establish and train a prediction model; Step 3: Determine the threshold for business decision-making based on model performance evaluation methods; Step 4: Input new data into the prediction model for prediction and determine the decision rule in combination with the threshold.

2. The method for market analysis and automated business decision-making according to claim 1, characterized in that, Step 1 also includes preprocessing the collected historical data.

3. The method for market analysis and automated business decision-making according to claim 2, wherein The preprocessing includes: Removing duplicate data, filling in missing values, correcting outliers, and unifying the data format.

4. The method for market analysis and automated business decision-making according to claim 1, characterized in that, The model performance evaluation method is the pb value or the F1 score.

5. The method for market analysis and automated business decision-making according to claim 1, characterized in that Step 4 also includes: Judging whether the input new data is emergency data. If it is emergency data, give priority to predicting the emergency data.

6. The method for market analysis and automated business decision-making according to claim 5, characterized in that, Step 1 also includes dividing the historical data into emergency data and non-emergency data, creating a data type recognition model based on the division result; Step 4 determines whether the input new data is emergency data based on the data type recognition model.

7. The method for market analysis and automated business decision-making according to claim 1, characterized in that It also includes Step 5: Regularly optimize the prediction model.

8. The method for market analysis and automated business decision-making according to claim 7, characterized in that, The optimization methods include: A. Monitor the decision execution situation and optimize the prediction model according to the monitoring results; and / or B. Introduce new data to retrain the prediction model.

9. A system for market analysis and automated business decision-making, characterized in that, Including: Data collection module: Obtain historical data related to target market analysis or business decision-making and their corresponding binary-type targets; Prediction model establishment module: Based on machine learning algorithms, use the sample objects and their corresponding attributes as inputs and the binary-type targets as outputs to establish and train a prediction model; Threshold determination module: Determine the threshold for business decision-making based on model performance evaluation methods; Prediction module: Obtain new data, input the new data into the prediction model for prediction, and determine the decision rule in combination with the threshold.

10. The system for market analysis and automated business decision-making according to claim 9, wherein It also includes: Data type recognition module: Used to determine whether new data is emergency data; When the new data is determined to be emergency data, give priority to predicting the emergency data.

Images