A customer demand forecasting method and system

Abstract

This application provides a customer demand forecasting method and system, applicable to the field of data processing technology. The method includes: extracting and analyzing features from multiple historical customer demand information to generate multiple historical customer demand feature information; classifying the multiple historical customer demand feature information to generate multiple historical customer demand group information; and performing forecasting calculations based on the multiple historical customer demand group information, randomly generated customer group demand forecasting weight information, randomly generated customer group demand forecasting bias information, preset customer group demand forecasting weight adjustment step size, and preset customer group demand forecasting bias adjustment step size to obtain multiple current customer demand information. This application dynamically optimizes forecasting parameters for different customer groups, enabling the customer demand forecasting results to effectively adapt to diverse computational resource constraints, improving forecasting accuracy and efficiency, thereby accurately grasping large-scale customer demand and facilitating the optimization of service resource allocation.

Description

Technical Field

[0001] This application belongs to the field of data processing technology, and in particular relates to methods and systems for predicting customer demand. Background Technology

[0002] Against the backdrop of the current digital wave sweeping across all industries, the field of customer demand forecasting is also undergoing continuous innovation. The integration of cutting-edge technologies such as big data and artificial intelligence is prompting companies to actively explore ways to uncover customer needs and improve service quality.

[0003] Current technologies typically rely on traditional data mining algorithms and basic machine learning models. For example, some companies use time series analysis to try to predict future demand trends based on the chronological order of historical customer demand data, such as estimating next month's sales by analyzing monthly product sales data from the past few years. Other companies use simple linear regression models to establish a linear relationship between some basic customer attributes (such as age, spending amount, etc.) and demand, thereby making preliminary quantitative predictions of customer demand.

[0004] However, existing time series analysis methods can only capture short-term, regular changes in data. They are difficult to effectively respond to sudden, irregular factors in the market, such as policy adjustments or sudden social events that cause drastic changes in customer demand. The prediction results often lag behind the actual changes in demand. Furthermore, simple linear regression models oversimplify the complex relationship between customer demand and influencing factors and cannot fully consider the diversity and dynamism of customer behavior. Summary of the Invention

[0005] In view of this, embodiments of this application provide a customer demand forecasting method and system, which aims to solve the problem of insufficient forecasting ability in the prior art when dealing with the complex and volatile nature of customer demand and the impact of sudden market factors.

[0006] A first aspect of this application provides a customer demand forecasting method, including:

[0007] Obtain information on multiple historical customer needs;

[0008] The multiple historical customer demand information is subjected to feature extraction and analysis to generate multiple historical customer demand feature information.

[0009] The multiple historical customer demand characteristic information is classified and processed to generate multiple historical customer demand group information;

[0010] Based on the information of multiple historical customer demand groups, multiple randomly generated customer demand prediction weights, multiple randomly generated customer demand prediction biases, preset customer demand prediction weight adjustment steps, and preset customer demand prediction bias adjustment steps, multiple current customer demand information is obtained through prediction calculation.

[0011] A second aspect of this application provides a customer demand forecasting system, comprising:

[0012] The historical customer demand information acquisition module is used to acquire multiple historical customer demand information.

[0013] The historical customer demand feature information generation module is used to extract and analyze the features of the multiple historical customer demand information to generate multiple historical customer demand feature information.

[0014] The historical customer demand group information generation module is used to classify and process the multiple historical customer demand feature information to generate multiple historical customer demand group information.

[0015] The current customer demand information generation module is used to perform prediction calculations based on the multiple historical customer demand group information, multiple randomly generated customer group demand prediction weight information, multiple randomly generated customer group demand prediction bias information, preset customer group demand prediction weight adjustment step size, and preset customer group demand prediction bias adjustment step size to obtain multiple current customer demand information.

[0016] A third aspect of this application provides a terminal device, the terminal device including a memory and a processor, the memory storing a computer program executable on the processor, the processor executing the computer program to implement the steps of the customer demand prediction method as described in the first aspect above.

[0017] A fourth aspect of this application provides a computer-readable storage medium, comprising: storing a computer program, which, when executed by a processor, implements the steps of the customer demand forecasting method described in the first aspect above.

[0018] The beneficial effects of this application's embodiments compared to the prior art are as follows: This application captures the key characteristics of customer needs at different granularities, groups customers with similar demand patterns into one category, deeply explores the unique demand patterns of different groups, and dynamically optimizes prediction parameters for the characteristics of different customer groups. This makes the calculated customer demand prediction results not only adaptable to diverse computing resource constraints, but also improve efficiency while ensuring prediction accuracy. This provides strong support for enterprises to accurately grasp the needs of large-scale customers and optimize service resource allocation, thereby achieving flexible and efficient demand prediction and avoiding resource waste or untimely demand response. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram illustrating the implementation process of the customer demand forecasting method provided in Embodiment 1 of this application;

[0021] Figure 2 This is a schematic diagram illustrating the implementation process of the customer demand forecasting method provided in Embodiment 2 of this application;

[0022] Figure 3 This is a schematic diagram illustrating the implementation process of the customer demand forecasting method provided in Embodiment 3 of this application;

[0023] Figure 4 This is a schematic diagram illustrating the implementation process of the customer demand forecasting method provided in Embodiment 4 of this application;

[0024] Figure 5 This is a schematic diagram illustrating the implementation process of the customer demand forecasting method provided in Embodiment 5 of this application;

[0025] Figure 6 This is a schematic diagram illustrating the implementation process of the customer demand forecasting method provided in Embodiment Six of this application;

[0026] Figure 7 This is a schematic diagram illustrating the implementation process of the customer demand forecasting method provided in Embodiment 7 of this application;

[0027] Figure 8 This is a schematic diagram of the customer demand forecasting system provided in the embodiments of this application;

[0028] Figure 9 This is a schematic diagram of the terminal device provided in the embodiments of this application. Detailed Implementation

[0029] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0030] To illustrate the technical solution described in this application, specific embodiments are provided below.

[0031] Figure 1 A flowchart illustrating the implementation of the customer demand forecasting method provided in Embodiment 1 of this application is shown, and is described in detail below:

[0032] Step S101: Obtain multiple historical customer demand information.

[0033] In this embodiment, historical customer demand information can be various demand-related records generated by customers during their interactions with the enterprise over a past period. These records include data reflecting customer demand characteristics, such as the content of the customer's demand for products or services, the time when the demand occurred, the frequency of the demand, and the scale of the demand. Historical customer demand information can be collected from relevant data in the enterprise's business systems. For example, customer order information (including the type, quantity, and time of the ordered products) can be extracted from the sales record system, customer inquiry and repair request records can be obtained from the customer service system, and customer browsing and click behavior data can be organized from the interaction logs. Subsequently, these collected data undergo preprocessing operations such as cleaning and standardization to remove invalid information and unify the data format, forming a set of historical customer demand information that can be used for subsequent feature extraction and analysis.

[0034] Step S102: Perform feature extraction and analysis on the multiple historical customer demand information to generate multiple historical customer demand feature information.

[0035] In this embodiment, multiple historical customer demand information sets can be collected and organized first. This information covers records of customers' past demands in terms of content, time, frequency, and scale, forming a raw dataset. Then, a deep neural network is used to preliminarily process this raw historical customer demand information, transforming it into a high-dimensional feature vector containing rich information. This vector can extract various potential information from the raw data. Subsequently, corresponding analyzers are set for different levels of feature granularity. These analyzers optimize sub-vectors of different lengths within the high-dimensional feature vector, ensuring that each sub-vector effectively captures the key characteristics of historical customer demands. During the optimization process, historical customer demand information can be combined with... The system uses user demand tagging information to calculate the deviation of the analysis results corresponding to each level of sub-vectors. These deviations are then integrated according to their importance. By continuously adjusting the network parameters, the sub-vectors at each level can accurately reflect the characteristics of historical customer demand. At the same time, information naturally diffuses to all dimensions during this process. Even intermediate-length sub-vectors that have not been specifically optimized can maintain good accuracy through this diffusion. Thus, for multiple input historical customer demand information, a high-dimensional vector containing features of different granularities can be generated through a single network processing. Each level of sub-vector is a valid historical customer demand feature information, thereby obtaining multiple historical customer demand feature information.

[0036] Step S103: Classify the multiple historical customer demand feature information to generate multiple historical customer demand group information.

[0037] In this embodiment, representative intermediate-granularity features from multiple historical customer demand feature information obtained through feature extraction can be selected as the classification basis. These features have captured the key commonalities and differences in customer demand through preprocessing. Then, several group cores are initialized, each representing a potential customer group. The similarity between each historical customer demand feature information and each group core is calculated, and the feature information is assigned to the most similar group. Then, the core of each group is updated according to the assignment results to make it more consistent with the average level of all feature information in the group. The steps of feature assignment and core update are repeated until the changes in the group cores are stable and no longer fluctuate significantly. After multiple iterations and optimizations, the originally scattered multiple historical customer demand feature information is divided into several groups with similar internal features and significant differences in external features. The set corresponding to each group is a historical customer demand group information, thereby generating multiple historical customer demand group information.

[0038] Step S104: Based on the multiple historical customer demand group information, multiple randomly generated customer group demand prediction weight information, multiple randomly generated customer group demand prediction bias information, preset customer group demand prediction weight adjustment step size, and preset customer group demand prediction bias adjustment step size, prediction calculation is performed to obtain multiple current customer demand information.

[0039] In this embodiment, the preset step size for adjusting the customer group demand prediction weights and the preset step size for adjusting the customer group demand prediction biases can be preset manually. It can be based on multiple historical customer demand group information sets, matching each group with corresponding randomly generated customer group demand prediction weights and biases. These weights and biases are used to initially construct the demand prediction relationship for each group. Then, using the actual demand data of historical customer demand groups, the deviation between the current prediction result and the actual result is calculated to evaluate the rationality of the weights and biases. Subsequently, according to the preset step size for adjusting the customer group demand prediction weights and the preset step size for adjusting the customer group demand prediction biases, the weights and biases with large deviations are adjusted: for groups with good prediction performance, small-scale fine adjustments are made near their weights and biases to further optimize prediction accuracy; for groups with poor prediction performance, the adjustment range is expanded to explore new combinations of weights and biases. The process of repeated deviation assessment and parameter adjustment, iteratively optimizing weights and biases, continues until the deviation between the predicted results and historical actual data stabilizes at a low level. Finally, the optimized weights and biases are applied to each historical customer demand group information, and the current customer demand information corresponding to each group is obtained through prediction calculation, thus generating multiple current customer demand information sets. The process of predicting using the optimized weights and biases can be as follows: First, for each historical customer demand group information, the corresponding optimized customer group demand prediction weight information and customer group demand prediction bias information are called. Then, the historical customer demand characteristic information of the group is associated and integrated with the corresponding weight information to capture the key influence relationship between features and demand. Next, the integration result is calibrated based on the bias information to obtain a preliminary basic value for demand prediction for the group. Then, referring to the historical demand change patterns and characteristic distribution characteristics within the group, the basic prediction value is fine-tuned to ensure that the prediction results closely match the actual demand patterns of the group. Finally, through the above integration, calibration, and fine-tuning steps, corresponding current customer demand information is generated for each historical customer demand group, thus obtaining multiple current customer demand information sets.

[0040] In this embodiment, the preset customer group demand prediction weight adjustment step size and the preset customer group demand prediction bias adjustment step size can be determined based on the feature dimensions and the number of groups in historical customer demand data. For example, if the dimension of historical customer demand feature information is 2048, and it is divided into 10 historical customer demand groups, the weight adjustment step size can be initially set between 0.01 and 0.1, and the bias adjustment step size can be set between 0.001 and 0.01. Then, the step size can be dynamically refined in combination with the degree of demand fluctuation within the group. For groups with large demand fluctuations (such as those where the demand standard deviation in historical data exceeds 30% of the mean), the adjustment step size can be appropriately increased (such as setting the weight step size to 0.05 and the bias step size to 0.005) to accelerate the adaptation to significant demand changes. For groups with stable demand (demand standard deviation below 10% of the mean), reduce the step size (e.g., set the weight step size to 0.02 and the bias step size to 0.002). Finally, step size calibration can be performed using small-scale validation data. Pre-training can be performed using some historical customer demand group information. If the prediction bias is found to be continuously decreasing and converging stably, maintain the current step size; if the bias fluctuates repeatedly, reduce the step size by 10%; if the convergence is too slow, increase the step size by 10%. Finally, determine the optimal adjustment step size combination suitable for different groups. For example, after calibration, the weight adjustment step size for a high-fluctuation group is set to 0.06 and the bias adjustment step size is set to 0.006, while for a low-fluctuation group, they are set to 0.015 and 0.0015 respectively, to achieve accurate and efficient parameter optimization.

[0041] The customer demand forecasting method provided in this application captures key characteristics of customer demand at different granularities, groups customers with similar demand patterns into one category, deeply explores the unique demand patterns of different groups, and dynamically optimizes forecasting parameters according to the characteristics of different customer groups. This ensures that the calculated customer demand forecasting results can not only adapt to diverse computing resource constraints, but also improve efficiency while ensuring forecasting accuracy. This provides strong support for enterprises to accurately grasp the needs of large-scale customers and optimize service resource allocation, thereby achieving flexible and efficient demand prediction and avoiding resource waste or untimely demand response.

[0042] Figure 2 The flowchart illustrating the implementation of the customer demand forecasting method provided in Embodiment 2 of this application is shown. The difference between this method and Embodiment 1 is that step S102 specifically includes:

[0043] Step S201: Encode and normalize the multiple historical customer demand information to generate multiple historical customer demand vectors.

[0044] In this embodiment, non-numerical data (such as demand content and product type) in historical customer demand information can be converted into numerical codes first, for example, encoding "hardware procurement" as 1 and "software service" as 2. Numerical data (such as demand quantity and consumption amount) can be normalized to the [0,1] range using the Min-Max method, for example, adjusting the quantity of a customer's single demand from 50 to 0.6 (the original range is 0-100). Subsequently, the encoded and normalized data are concatenated according to time order or feature category to generate a fixed-length historical customer demand vector, such as a 256-dimensional vector containing dimensions such as demand type, timestamp, quantity, and amount.

[0045] Step S202: Based on the preset historical customer demand query feature extraction vector, perform feature extraction calculation on the multiple historical customer demand vectors to obtain multiple historical customer demand query feature information.

[0046] In this embodiment, the preset historical customer demand query feature extraction vector can be manually preset or a randomly initialized 128-dimensional vector. Feature extraction is achieved by performing matrix multiplication with the historical customer demand vector. For example, multiplying the 256-dimensional historical customer demand vector with the 128-dimensional query extraction vector yields 128-dimensional historical customer demand query feature information, focusing on the core objectives of customer demand (such as features related to "whether it is an urgent demand"). The initial value range of the query extraction vector is set to [-0.1, 0.1], and subsequent fine-tuning is performed through training to adapt to the demand prediction task.

[0047] Step S203: Based on the preset historical customer demand key feature extraction vector, perform feature extraction calculation on the multiple historical customer demand vectors to obtain multiple historical customer demand key feature information.

[0048] In this embodiment, the preset historical customer demand key feature extraction vector can be manually set and can be three sets of 128-dimensional random vectors (corresponding to three sub-features: demand time, type, and scale). Each set of vectors is multiplied by the corresponding dimension sub-vector in the historical customer demand vector. The multiplication result serves as the historical customer demand key feature information to extract key attribute features. For example, multiplying the time dimension sub-vector by the first set of key vectors yields key features reflecting the periodicity of demand; multiplying the type dimension sub-vector by the second set of key vectors yields key features distinguishing demand categories. The initial range of the key extraction vector is set to [-0.1, 0.1], and the weights of different sets of vectors are adjusted independently.

[0049] Step S204: Based on the preset historical customer demand value feature extraction vector, perform feature extraction calculation on the multiple historical customer demand vectors to obtain multiple historical customer demand value feature information.

[0050] In this embodiment, the preset historical customer demand value feature extraction vector can be manually preset and can be a 128-dimensional vector matching the dimension of the key feature extraction vector. Through linear transformation, numerical attributes of specific demand features are extracted from the historical customer demand vector, such as the fluctuation range of demand quantity and the distribution characteristics of service duration. The initial construction of the value extraction vector references the feature importance of historical data, assigning higher initial weights (e.g., 0.2) to high-frequency demand features (e.g., "demand quantity") and lower initial weights (e.g., 0.05) to low-frequency features (e.g., "remarks information"). Alternatively, the preset historical customer demand value feature extraction vector can be multiplied by multiple historical customer demand vectors, with the multiplication result serving as multiple historical customer demand value feature information.

[0051] Step S205: Based on the multiple historical customer demand query feature information and the multiple historical customer demand key feature information, perform interactive processing to obtain multiple historical customer demand interaction information.

[0052] In this embodiment, the similarity between each historical customer demand query feature and all historical customer demand key feature information is calculated. For example, the association strength is measured by vector dot product. The dot product result of each historical customer demand query feature and all historical customer demand key feature information can be normalized by the softmax function to obtain historical customer demand interaction information. The interaction results corresponding to the top 5 key features with the highest historical customer demand interaction information are concatenated to generate historical customer demand interaction information, capturing the matching relationship between the query target and key features, such as the strong interaction between the "urgent need" query and the "time urgency" key feature.

[0053] Step S206: Based on the multiple historical customer demand value feature information and multiple historical customer demand interaction information, perform fusion processing to generate multiple historical customer demand feature variables.

[0054] In this embodiment, the interaction results corresponding to the top 5 key features with the highest historical customer demand interaction information can be used as coefficients to perform a weighted summation of the corresponding historical customer demand value feature information. For example, the "quantity fluctuation" value feature with a weight of 0.3 and the "time distribution" value feature with a weight of 0.2 can be weighted and fused. After fusion, it can be converted into a 64-dimensional vector through a fully connected layer, which is the historical customer demand feature variable, integrating the numerical attributes and interaction importance of key features.

[0055] Step S207: Based on the multiple historical customer demand feature variables, perform feature transformation and enhancement processing on the multiple historical customer demand vectors to generate multiple historical customer demand feature information.

[0056] In this embodiment, historical customer demand feature variables and the original historical customer demand vector can be concatenated dimensionally, and the original information can be preserved through residual connection. Then, a nonlinear transformation is introduced using the ReLU function to enhance the feature representation capability. The content calculated by the ReLU function is used as the historical customer demand feature information. For example, a 256-dimensional original vector and a 64-dimensional feature variable can be concatenated into a 320-dimensional vector. After transformation, historical customer demand feature information containing multi-scale features is generated, covering original attributes and interactive enhancement features.

[0057] The customer demand prediction method provided in this application further enhances the richness and relevance of historical customer demand features through refined feature encoding, multi-dimensional extraction and interactive fusion, providing a more accurate feature foundation for subsequent group classification and demand prediction. While ensuring efficiency, it enhances the ability to capture complex customer demand patterns and helps enterprises grasp customer demand patterns more accurately.

[0058] Figure 3 The flowchart illustrating the implementation of the customer demand forecasting method provided in Embodiment 3 of this application is shown. The difference between this method and Embodiment 1 is that step S102 specifically includes:

[0059] Step S301: Based on the preset historical customer demand frequency feature extraction vector, perform feature extraction processing on the multiple historical customer demand information to obtain multiple historical customer demand frequency feature information.

[0060] In this embodiment, the preset historical customer demand frequency feature extraction vector can be manually set. It can be based on frequency-related dimensions such as "weekly demand frequency," "demand interval in days," and "quarterly peak demand frequency" in historical data, initializing a 16-dimensional weighted vector. The "weekly demand frequency" dimension, which has a significant impact on high-frequency demand, is weighted at 0.3, the "quarterly peak demand frequency" dimension, which reflects long-term patterns, is weighted at 0.25, and the weights of the remaining dimensions are distributed in descending order of importance (e.g., 0.15, 0.1, etc.). By weighting the frequency-related data in the historical customer demand information (e.g., a customer "demands twice a week, with a demand interval of 3 days") with this vector, features that reflect the pattern of demand occurrence are extracted, generating multiple historical customer demand frequency feature information.

[0061] Step S302: Based on the preset historical customer demand spatiotemporal preference feature extraction vector, perform feature extraction processing on the multiple historical customer demand information to obtain multiple historical customer demand spatiotemporal preference feature information.

[0062] In this embodiment, the preset historical customer demand spatiotemporal preference feature extraction vector can be manually set. It can be initialized with a 20-dimensional weight vector based on the customer's time preference (such as "weekday / weekend" and "peak season / off-season") and spatial attributes (such as "service area" and "remote / on-site demand"). In the time preference dimension, the weight of "peak season demand ratio" is set to 0.3, and the weight of "weekend demand frequency" is set to 0.2. In the spatial attribute dimension, the weight of "core area demand ratio" is set to 0.25, and the weights of the other dimensions are set to 0.05~0.1. For example, the information that a customer's "peak season demand ratio is 70% and core area demand ratio is 80%" can be weighted and extracted to generate features reflecting spatiotemporal preferences, resulting in multiple historical customer demand spatiotemporal preference feature information. This can adapt to the logic of capturing information of different dimensions with multi-granularity features.

[0063] Step S303: Based on the preset historical customer demand correlation feature extraction vector, perform feature extraction processing on the multiple historical customer demand information to obtain multiple historical customer demand correlation feature information.

[0064] In this embodiment, the preset historical customer demand correlation feature extraction vector can be manually set, focusing on the dependencies between demands (such as "whether hardware purchase is accompanied by installation service" or "the frequency of software upgrades and technical support"). A 24-dimensional weight vector is initialized. The weight of the dimension corresponding to strongly correlated demand combinations (such as "hardware purchase - installation service") can be set to 0.35, the weight of the dimension corresponding to weakly correlated but potentially important dimensions (such as "demand conversion rate within 7 days after consultation") can be set to 0.2, and the weight of other correlated dimensions can be set to 0.05~0.15. By weighting the related events recorded in historical customer demands, the synergistic patterns between demands are extracted, generating multiple historical customer demand correlation feature information.

[0065] Step S304: Match the frequency characteristics of the occurrence of multiple historical customer demands with the spatiotemporal preference characteristics of multiple historical customer demands to obtain the spatiotemporal preference frequency matching information of multiple historical customer demands.

[0066] In this embodiment, the degree of matching between the frequency characteristics of each historical customer demand and the spatiotemporal preference characteristics of historical customer demand can be calculated. For example, by comparing the overlap between "high-frequency demand during peak season" and "peak season periods in spatiotemporal preference," high weights are assigned to feature combinations with high matching degrees (e.g., a weight of 0.8 is set for overlap of over 80%), and low weights are assigned to combinations with low matching degrees (e.g., a weight of 0.2 is set for overlap of less than 30%). The weighted matching results are then split according to time periods (day, week, month) to generate multiple historical customer demand spatiotemporal preference frequency matching information, capturing the pattern of "when and where demand occurs frequently." This can be achieved by nesting features to associate information of different granularities.

[0067] Step S305: Based on the interrelationship feature information of the multiple historical customer needs and the spatiotemporal preference frequency matching information of the multiple historical customer needs, perform fusion processing to generate multiple historical customer need feature information.

[0068] In this embodiment, historical customer demand correlation features and spatiotemporal preference frequency matching information can be aligned by dimension. Key correlation features (such as "high-frequency hardware demand during peak season accompanied by installation services") are assigned a fusion weight of 0.4, core spatiotemporal frequency features (such as "core regional demand during quarterly peak periods") are assigned a weight of 0.3, and the remaining features are assigned the remaining weights according to their importance. By integrating the correlation patterns and spatiotemporal frequency patterns of demand through weighted fusion, and then performing multi-layer processing to retain multi-granular features (such as macro-level quarterly patterns and micro-level weekly fluctuations), multiple historical customer demand feature information are finally generated, achieving a comprehensive characterization of demand in terms of frequency, spatiotemporal, and correlation dimensions.

[0069] The customer demand prediction method provided in this application focuses on three core dimensions: demand frequency, spatiotemporal preference, and correlation. By combining precise feature extraction and multi-dimensional fusion with preset extraction vectors, the generated historical customer demand feature information can more comprehensively capture the dynamic patterns and internal correlations of customer demand. This provides a more realistic feature foundation for subsequent group classification and demand prediction, helping enterprises to achieve more accurate pattern mining and predictive decision-making in large-scale customer demand analysis.

[0070] Figure 4 The flowchart illustrating the implementation of the customer demand forecasting method provided in Embodiment 4 of this application is shown. The difference between this method and Embodiment 1 is that step S103 specifically includes:

[0071] Step S401: Based on the preset historical customer demand group quantity information, randomly extract the multiple historical customer demand feature information to obtain multiple historical customer demand central feature information.

[0072] In this embodiment, the preset number of historical customer demand groups can be manually set, based on the total number and diversity of historical customer demand feature information. If the total number of feature information is 1000 and the demand patterns differ significantly, the preset number of groups can be 5-8; if the feature similarity is high, the preset number can be 3-5. For example, for a dataset containing 1200 feature information, the preset number of groups is 6. Subsequently, feature information equal to the number of groups is randomly extracted from multiple historical customer demand feature information sets as initial cores. For example, 6 features are randomly selected from 1200, each representing the initial core of a group, generating multiple historical customer demand core feature information sets.

[0073] Step S402: Based on the multiple historical customer demand feature information and the multiple historical customer demand central feature information, multiple historical customer demand peripheral feature information are obtained.

[0074] In this embodiment, the remaining historical customer demand features, excluding those already extracted as central features, can be defined as peripheral features. For example, out of 1200 features, 6 are central features, and the remaining 1194 are peripheral features, ensuring that each peripheral feature can be used for subsequent distance calculations with the central feature.

[0075] Step S403: Calculate the Manhattan distance between the central feature information of the multiple historical customer demands and the peripheral feature information of the multiple historical customer demands to obtain the radial distance information of the multiple historical customer demand features; the radial distance information of the historical customer demand features corresponds one-to-one with the central feature information of the historical customer demands and the peripheral feature information of the historical customer demands.

[0076] In this embodiment, the Manhattan distance can be calculated separately for each historical customer demand peripheral feature information and all central feature information. This involves calculating the absolute value of the difference along each feature dimension and summing them. For example, if the absolute values ​​of the differences between a peripheral feature and a central feature in five dimensions ("demand frequency," "spatiotemporal preference," etc.) are 0.2, 0.3, 0.1, 0.4, and 0.2 respectively, the summed radial distance is 1.2. The distance results between each peripheral feature and each central feature are recorded separately, generating multiple historical customer demand feature radial distance information that correspond one-to-one, quantifying the degree of difference between features.

[0077] Step S404: The minimum value of the radial distance information of the multiple historical customer demand features corresponding to the multiple historical customer demand surrounding feature information is taken as the group confirmation distance information of the multiple historical customer demand surrounding feature information.

[0078] In this embodiment, for each historical customer demand peripheral feature information, the minimum value is selected from the radial distances between it and all central features, and this minimum value is used as the basis for confirming that the peripheral feature belongs to a certain group. For example, if the distances between a peripheral feature and six central features are 1.2, 0.8, 1.5, 0.9, 1.1, and 1.3, the minimum value of 0.8 is the group confirmation distance information, reflecting the correlation strength between the peripheral feature and the nearest central feature.

[0079] Step S405: Based on the distance information of the surrounding characteristic groups of the multiple historical customer needs and the central characteristic information of the historical customer needs, classify the surrounding characteristic information of the multiple historical customer needs to generate multiple historical customer needs to be confirmed group information.

[0080] In this embodiment, the peripheral feature information of each historical customer demand can be assigned to the group containing the central feature corresponding to its confirmation distance. For example, peripheral features with a minimum distance of 0.8 are assigned to the group corresponding to the central feature. After all peripheral features are assigned, a temporary group is formed with the central feature as the core and the peripheral features as members. Each group contains the central feature and the associated peripheral features, generating multiple groups of historical customer demand information to be confirmed. If a group contains less than 5% of the total number of peripheral features (e.g., a group contains only 20 out of 1200), it can be marked as a group to be adjusted.

[0081] Step S406: Iteratively classify the information of the multiple historical customer demand groups to be confirmed to obtain the surrounding feature information of multiple historical customer demand groups.

[0082] In this embodiment, the average feature value of all feature information within each group to be confirmed can be recalculated and used as the new central feature. Steps S403-S405 are repeated to recalculate the distance and assign surrounding features according to the new central feature until the change in the group central feature between two iterations is less than a preset threshold (e.g., the average difference of each dimension < 0.05). For example, after iteration, the "demand frequency" dimension is adjusted from 0.6 to 0.58 and the "spatiotemporal preference" dimension is adjusted from 0.7 to 0.69. The iteration stops when the changes stabilize. Each group ultimately formed contains stable central features and attribution features, generating multiple historical customer demand group information to ensure that features are similar within groups and that differences are significant between groups.

[0083] The customer demand prediction method provided in this application, through preset reasonable group size, distance-based accurate classification and iterative optimization, generates historical customer demand group information that can more accurately reflect the demand patterns of different customer groups, providing a more targeted group basis for subsequent prediction, improving the rationality of group division while ensuring classification efficiency, and helping enterprises accurately discover the demand patterns of different groups.

[0084] Figure 5 The flowchart illustrating the implementation of the customer demand forecasting method provided in Embodiment 5 of this application is shown. The difference between this method and Embodiment 4 above is that step S406 specifically includes:

[0085] Step S501: Calculate the mean of the multiple historical customer demand confirmation groups to obtain the central information of the multiple historical customer demand confirmation groups.

[0086] In this embodiment, the average value of all feature information contained in each historical customer demand confirmation group information is calculated by dimension. For example, a certain group has feature values ​​of 0.6, 0.7, and 0.5 in the "demand frequency" dimension, with an average of 0.6; and feature values ​​of 0.8, 0.7, and 0.9 in the "spatiotemporal preference" dimension, with an average of 0.8. The average values ​​of each dimension are combined to form a new central pivot for the group, generating multiple central pivot information for historical customer demand confirmation groups. The new central pivot needs to conform to the overall distribution of features within the group.

[0087] Step S502: Determine whether the central information of the multiple historical customer demand groups to be confirmed is the same as the central feature information of multiple historical customer demand groups; if yes, proceed to step S503; if no, proceed to step S504.

[0088] In this embodiment, the preset judgment threshold is that the sum of the absolute values ​​of the differences between the features of each dimension is ≤0.1. For example, comparing the difference of "demand frequency" between the new center and the original center (0.02), the difference of "spatiotemporal preference" (0.03), and the total difference (0.05 ≤ 0.1), the new center is judged to be the same. If the total difference is >0.1 (e.g., the sum of the differences between the original center and the new center is 0.15), the new center is judged to be different. This threshold is set based on a reasonable range of feature fluctuations in historical data to ensure that the iteration is terminated only after the group center has stabilized, avoiding premature termination of classification due to small fluctuations.

[0089] Step S503: Generate multiple historical customer demand group information based on the multiple historical customer demand group information to be confirmed.

[0090] In this embodiment, when the central characteristic of the group to be confirmed is the same as the original central characteristic, it indicates that the core of the group has stabilized. At this time, each group to be confirmed is directly identified as the final group, which includes stable central characteristics and surrounding characteristic information. For example, after the central characteristic of 6 groups to be confirmed is determined to be stable, the core dimension characteristics of each group, such as "demand frequency" and "spatiotemporal preference", no longer change significantly, generating multiple historical customer demand group information to ensure high similarity of characteristics within the group and obvious differences between groups.

[0091] Step S504: The central information of the multiple historical customer demand groups to be confirmed is used as the central feature information of multiple historical customer demand groups, and the process returns to step S402.

[0092] In this embodiment, if the central hub of the group to be confirmed is different from the original central hub, it means that the group core still needs to be optimized. The newly calculated central hub information of the group to be confirmed replaces the original central hub feature information. For example, the "demand frequency" central hub with a mean of 0.6 replaces the original central hub's 0.5 as the initial central hub for the new round of iteration. Then, the distance between the surrounding features and the new central hub is recalculated and the group is allocated until the central hub is stable.

[0093] The customer demand prediction method provided in this application, through a clear mean calculation and threshold judgment mechanism, achieves dynamic optimization of the group center and precise control of iteration termination, avoiding blind iteration or premature convergence in the classification process. The generated historical customer demand group information is more stable and has stronger feature representativeness, providing more reliable customer group information for subsequent demand prediction and helping enterprises to more accurately explore the demand patterns of different groups.

[0094] Figure 6 The flowchart illustrating the implementation of the customer demand forecasting method provided in Embodiment Six of this application is shown. The difference between this method and Embodiment One is that step S104 specifically includes:

[0095] Step S601: Generate multiple customer group demand prediction parameter group information based on the multiple randomly generated customer group demand prediction weight information and the multiple randomly generated customer group demand prediction bias information; the customer group demand prediction parameter group information includes customer group demand prediction weight information and customer group demand prediction bias information.

[0096] In this embodiment, the weight information for predicting customer demand for each randomly generated customer group can be combined with the corresponding randomly generated bias information for predicting customer demand for the same customer group to form a parameter set. For example, random weights (such as 0.2, 0.3, 0.1, etc.) and biases (such as 0.05, 0.03, 0.02, etc.) can be generated for six historical customer demand groups. The weight and bias of each group correspond one-to-one, generating six customer demand prediction parameter sets to ensure that each group has an independent combination of prediction parameters.

[0097] Step S602: Calculate the multiple customer group demand prediction accuracy characterization information based on the multiple historical customer demand group information and the multiple customer group demand prediction parameter group information.

[0098] In this embodiment, the demand prediction parameter group information for each customer group can be applied to the corresponding historical customer demand group information. The group characteristics are weighted and integrated through the weights in the parameter group, and the prediction result is obtained by combining the bias calibration. Then, the prediction result is compared with the actual historical demand data of the group, and the degree of deviation (such as mean absolute error) is calculated as an accuracy characterization. For example, if the average error between the prediction result and the actual demand of a certain group is 5%, then its accuracy characterization information is 95% (1-error rate). Multiple customer group demand prediction accuracy characterization information is generated to quantify the prediction effect of the parameter group.

[0099] Step S603: Determine whether the maximum value of the multiple customer group demand prediction accuracy representation information is greater than or equal to the preset customer group demand prediction accuracy representation threshold; if yes, proceed to step S604; if no, proceed to step S605.

[0100] In this embodiment, the preset threshold for customer group demand prediction accuracy can be set manually, based on business needs. For example, it could be set to 90% for core customer groups and 85% for general groups, resulting in a combined threshold of 88%. If the maximum value of the accuracy representation information is 92% ≥ 88%, the current parameter group is considered to meet the standard; if the maximum value is 85% < 88%, it is considered not to meet the standard. This threshold references the average accuracy of historical prediction tasks to ensure that the prediction results meet the actual application requirements and avoid low accuracy affecting decision-making.

[0101] Step S604: Based on the customer group demand prediction parameter group information corresponding to the maximum value of the multiple customer group demand prediction accuracy characterization information and multiple historical customer demand group information, prediction calculation is performed to obtain multiple current customer demand information.

[0102] In this embodiment, when the maximum accuracy is achieved, the parameter set corresponding to that maximum value is selected as the optimal parameter set and applied to all historical customer demand group information. By integrating group characteristics with optimal weights, combined with bias calibration, and fine-tuned by referring to the historical demand patterns of the groups, the current demand prediction results for each group are generated, such as "the current demand of the high-frequency peak season group is expected to increase by 10%" and "the demand of the stable area group remains stable," etc., thus obtaining multiple current customer demand information and ensuring that the prediction results are generated based on the optimal parameters.

[0103] Step S605: Based on the multiple customer group demand prediction accuracy characterization information, multiple customer group demand prediction parameter group information, preset customer group demand prediction weight adjustment step size, and preset customer group demand prediction bias adjustment step size, optimization calculation is performed to generate multiple customer group demand prediction parameter group information to be optimized.

[0104] In this embodiment, for parameter groups whose accuracy does not meet the standard, the parameters are adjusted according to the accuracy characterization information: the parameter group with low accuracy (such as 80% accuracy) is adjusted by expanding the adjustment range by a preset step size (weight step size 0.05, bias step size 0.005), and the parameter group with accuracy close to the threshold (such as 86% accuracy) is finely adjusted by a small step size (weight step size 0.02, bias step size 0.002). For example, a certain weight is increased from 0.2 to 0.25, and the bias is finely adjusted from 0.05 to 0.045, generating multiple customer group demand prediction parameter group information to be optimized.

[0105] Step S606: The information of the multiple customer group demand prediction parameter group to be optimized is taken as the information of multiple customer group demand prediction parameter group, and the process is returned to step S602.

[0106] In this embodiment, if the accuracy is not met, the optimized parameter set replaces the original parameter set, and the iteration is repeated until the maximum accuracy is achieved. For example, after two iterations, the accuracy of the parameter set increases from 85% to 90%, and the iteration stops after meeting the threshold requirement, ensuring that a high-quality parameter set is obtained through continuous optimization.

[0107] The customer demand forecasting method provided in this application provides a closed-loop mechanism of parameter group generation, accuracy evaluation, and dynamic optimization to ensure that the forecast parameters always adapt to the characteristics of the group's demand. Under the premise of meeting the accuracy threshold, it generates reliable current customer demand information, which not only improves the pertinence of parameter optimization but also ensures the practicality of the forecast results, helping enterprises to achieve efficient resource allocation and rapid demand response based on accurate forecasting.

[0108] Figure 7 The flowchart illustrating the implementation of the customer demand forecasting method provided in Embodiment Seven of this application is shown. The difference between this method and Embodiment Six is ​​that step S605 specifically includes:

[0109] Step S701: The customer group demand prediction parameter group information corresponding to the maximum value of the multiple customer group demand prediction accuracy characterization information is taken as the core parameter group information for customer group demand prediction.

[0110] In this embodiment, the maximum value can be selected from the demand prediction accuracy representation information of multiple customer groups. For example, if the maximum accuracy is 92%, the parameter set corresponding to this maximum value (such as weight 0.3 and bias 0.04) can be determined as the core parameter set as the benchmark for optimization. The core parameter set should have the best prediction effect at present, so as to provide a reference for subsequent optimization of peripheral parameters.

[0111] Step S702: Based on the multiple customer group demand prediction parameter group information and the customer group demand prediction core parameter group information, obtain multiple customer group demand prediction peripheral parameter group information.

[0112] In this embodiment, the parameter groups other than the core parameter group can be defined as peripheral parameter groups. For example, if there are a total of 6 parameter groups, one of which is the core parameter group, then the remaining 5 are peripheral parameter groups. The peripheral parameter groups need to be compared with the core parameter group, and their accuracy representation information is lower than that of the core parameter group (such as 85%, 88%, etc.), providing candidate objects to be adjusted for parameter optimization.

[0113] Step S703: Based on the core parameter group information for customer group demand prediction, the preset customer group demand prediction weight adjustment step size, and the preset customer group demand prediction bias adjustment step size, optimize the peripheral parameter group information for multiple customer group demand predictions to obtain peripheral parameter group information for multiple customer group demand predictions to be optimized.

[0114] In this embodiment, the preset step size for adjusting the customer group demand prediction weights and the step size for adjusting the bias can be set based on the difference between the accuracy of the core parameter group and the peripheral parameter group: if the difference between the peripheral parameter group accuracy and the core accuracy is large (e.g., difference ≥ 5%), a larger step size is used (weight step size 0.05, bias step size 0.005); if the difference is small (e.g., difference < 3%), a smaller step size is used (weight step size 0.02, bias step size 0.002). For example, if the accuracy of a certain peripheral parameter group is 85%, which is 7% different from the core accuracy of 92%, the weight is adjusted from 0.2 to 0.25 and the bias is adjusted from 0.03 to 0.035 with a large step size, generating a peripheral parameter group to be optimized, so that the peripheral parameters are closer to the core parameters.

[0115] Step S704: Based on the core parameter group information for customer group demand prediction and the peripheral parameter group information for customer group demand prediction to be optimized, generate multiple parameter groups for customer group demand prediction to be optimized.

[0116] In this embodiment, the core parameter set can be combined with the optimized peripheral parameter set to form a new set of parameter sets to be optimized. For example, the core parameter set (92% accuracy) can be retained, and the five optimized peripheral parameter sets (accuracy improved to 88%~90%) can be included to generate six parameter sets to be optimized. The parameter sets to be optimized need to cover the stability advantages of the core parameters and the optimization potential of the peripheral parameters, so as to provide better candidate parameters for the next round of accuracy evaluation.

[0117] The customer demand forecasting method provided in this application uses a hierarchical optimization mechanism that determines the optimization benchmark through a core parameter set and makes directional adjustments to the peripheral parameter set. This retains the advantages of the current optimal parameters while using dynamic step sizes to drive the peripheral parameters to converge toward higher precision, reducing the cost of blind iteration. This makes parameter optimization more targeted and efficient, and the generated parameter set to be optimized can reach the accuracy threshold more quickly. This provides strong support for the final generation of reliable current customer demand information and helps enterprises improve the accuracy and efficiency of demand forecasting.

[0118] Corresponding to the method in the above embodiments, Figure 8 A structural block diagram of the customer demand forecasting system provided in the embodiments of this application is shown. For ease of explanation, only the parts related to the embodiments of this application are shown. Figure 8 The example customer demand forecasting system can be the execution entity of the customer demand forecasting method provided in the aforementioned embodiment 1.

[0119] Reference Figure 8 The customer demand forecasting system includes:

[0120] The historical customer demand information acquisition module 810 is used to acquire multiple historical customer demand information.

[0121] The historical customer demand feature information generation module 820 is used to extract and analyze the features of the multiple historical customer demand information to generate multiple historical customer demand feature information.

[0122] The historical customer demand group information generation module 830 is used to classify and process the multiple historical customer demand feature information to generate multiple historical customer demand group information.

[0123] The current customer demand information generation module 840 is used to perform prediction calculations based on the multiple historical customer demand group information, multiple randomly generated customer group demand prediction weight information, multiple randomly generated customer group demand prediction bias information, preset customer group demand prediction weight adjustment step size, and preset customer group demand prediction bias adjustment step size to obtain multiple current customer demand information.

[0124] The process by which each module in the customer demand forecasting system provided in this application implements its respective function can be found in the foregoing. Figure 1 The description of Embodiment 1 shown will not be repeated here.

[0125] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0126] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0127] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0128] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0129] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance. It should also be understood that although the terms "first," "second," etc., are used in the text to describe various elements in some embodiments of this application, these elements should not be limited by these terms. These terms are merely used to distinguish one element from another. For example, a first table may be named a second table, and similarly, a second table may be named a first table, without departing from the scope of the various described embodiments. Both the first table and the second table are tables, but they are not the same table.

[0130] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0131] The customer demand prediction method provided in this application can be applied to terminal devices such as mobile phones, tablets, wearable devices, in-vehicle devices, augmented reality (AR) / virtual reality (VR) devices, laptops, ultra-mobile personal computers (UMPCs), netbooks, and personal digital assistants (PDAs). This application does not impose any restrictions on the specific type of terminal device.

[0132] For example, the terminal device may be a station (STAION, ST) in a WLAN, a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a Wireless Local Loop (WLL) station, a Personal Digital Assistant (PDA) device, a handheld device with wireless communication capabilities, a computing device or other processing device connected to a wireless modem, an in-vehicle device, a vehicle networking terminal, a computer, a laptop computer, a handheld communication device, a handheld computing device, a satellite wireless device, a wireless modem card, a set-top box (STB), customer premises equipment (CPE), and / or other devices used for communication over a wireless system, as well as next-generation communication systems, such as mobile terminals in 5G networks or mobile terminals in future evolved Public Land Mobile Network (PLMN) networks.

[0133] As an example and not a limitation, when the terminal device is a wearable device, the term "wearable device" can also refer to any device that utilizes wearable technology to intelligently design and develop everyday wearables, such as glasses, gloves, watches, clothing, and shoes. Wearable devices are portable devices worn directly on the body or integrated into a user's clothing or accessories. Wearable devices are not merely hardware devices; they achieve powerful functions through software support, data interaction, and cloud interaction. Broadly defined, wearable smart devices include those with comprehensive functions, large sizes, and the ability to perform complete or partial functions without relying on a smartphone, such as smartwatches or smart glasses, as well as those focused on a specific application function that require interaction with other devices such as smartphones, such as various smart bracelets and smart jewelry for vital sign monitoring.

[0134] Figure 9 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. For example... Figure 9 As shown, the terminal device 9 of this embodiment includes: at least one processor 90 ( Figure 9 (Only one is shown in the image), memory 91, which stores a computer program 92 that can run on the processor 90. When the processor 90 executes the computer program 92, it implements the steps in the various customer demand forecasting method embodiments described above, for example... Figure 1 Steps S101 to S104 are shown. Alternatively, when the processor 90 executes the computer program 92, it implements the functions of each module / unit in the above system embodiments, for example... Figure 8The functions of modules 810 to 840 are shown.

[0135] The terminal device 9 can be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor 90 and a memory 91. Those skilled in the art will understand that... Figure 9 This is merely an example of terminal device 9 and does not constitute a limitation on terminal device 9. It may include more or fewer components than shown, or combine certain components, or different components. For example, the terminal device may also include input transmission devices, network access devices, buses, etc.

[0136] The processor 90 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0137] In some embodiments, the memory 91 may be an internal storage unit of the terminal device 9, such as a hard disk or memory of the terminal device 9. The memory 91 may also be an external storage device of the terminal device 9, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal device 9. Furthermore, the memory 91 may include both internal and external storage units of the terminal device 9. The memory 91 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory 91 can also be used to temporarily store data that has been sent or will be sent.

[0138] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0139] This application also provides a terminal device, which includes at least one memory, at least one processor, and a computer program stored in the at least one memory and executable on the at least one processor. When the processor executes the computer program, it causes the terminal device to implement the steps in any of the above method embodiments.

[0140] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0141] This application provides a computer program product that, when run on a terminal device, enables the terminal device to implement the steps described in the above-described method embodiments.

[0142] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0143] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0144] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0145] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0146] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method of predicting customer demand, characterized by, include: Obtain information on multiple historical customer needs; The multiple historical customer demand information is subjected to feature extraction and analysis to generate multiple historical customer demand feature information. The multiple historical customer demand characteristic information is classified and processed to generate multiple historical customer demand group information; Based on the multiple randomly generated customer group demand prediction weight information and the multiple randomly generated customer group demand prediction bias information, multiple customer group demand prediction parameter group information is generated; the customer group demand prediction parameter group information includes customer group demand prediction weight information and customer group demand prediction bias information. Based on the information of multiple historical customer demand groups and the information of multiple customer group demand prediction parameter groups, calculate the demand prediction accuracy representation information of multiple customer groups. Determine whether the maximum value of the multiple customer group demand prediction accuracy representation information is greater than or equal to the preset customer group demand prediction accuracy representation threshold. If so, then based on the customer group demand prediction parameter group information corresponding to the maximum value of the multiple customer group demand prediction accuracy characterization information and multiple historical customer demand group information, prediction calculation is performed to obtain multiple current customer demand information. If not, the customer group demand prediction parameter group information corresponding to the maximum value of the multiple customer group demand prediction accuracy characterization information shall be used as the core parameter group information for customer group demand prediction. Based on the information of the multiple customer group demand prediction parameter groups and the information of the core customer group demand prediction parameter groups, we obtain information of the multiple customer group demand prediction peripheral parameter groups. Based on the core parameter group information for customer group demand prediction, the preset customer group demand prediction weight adjustment step size, and the preset customer group demand prediction bias adjustment step size, the peripheral parameter group information for multiple customer group demand predictions is optimized to obtain multiple peripheral parameter group information for customer group demand predictions to be optimized. Based on the core parameter group information for customer group demand prediction and the peripheral parameter group information for customer group demand prediction to be optimized, multiple parameter groups for customer group demand prediction to be optimized are generated. The information of the multiple customer group demand prediction parameter groups to be optimized is used as the multiple customer group demand prediction parameter group information, and the process is returned to the step of calculating the multiple customer group demand prediction accuracy characterization information based on the multiple historical customer demand group information and the multiple customer group demand prediction parameter group information.

2. The customer demand forecasting method as described in claim 1, characterized in that, The step of extracting and analyzing features from the multiple historical customer demand information to generate multiple historical customer demand feature information specifically includes: The multiple historical customer demand information is encoded and normalized to generate multiple historical customer demand vectors; Based on the preset historical customer demand query feature extraction vector, feature extraction calculation is performed on the multiple historical customer demand vectors to obtain multiple historical customer demand query feature information. Based on the preset historical customer demand key feature extraction vector, feature extraction calculation is performed on the multiple historical customer demand vectors to obtain multiple historical customer demand key feature information. Based on the preset historical customer demand value feature extraction vector, feature extraction calculation is performed on the multiple historical customer demand vectors to obtain multiple historical customer demand value feature information. Based on the interactive processing of the multiple historical customer demand query feature information and the multiple historical customer demand key feature information, multiple historical customer demand interaction information are obtained. The multiple historical customer demand value feature information and multiple historical customer demand interaction information are fused together to generate multiple historical customer demand feature variables. Based on the multiple historical customer demand feature variables, feature transformation and enhancement processing are performed on the multiple historical customer demand vectors to generate multiple historical customer demand feature information.

3. The customer demand forecasting method as described in claim 1, characterized in that, The step of extracting and analyzing features from the multiple historical customer demand information to generate multiple historical customer demand feature information specifically includes: Based on the preset historical customer demand frequency feature extraction vector, feature extraction processing is performed on the multiple historical customer demand information to obtain multiple historical customer demand frequency feature information. Based on the preset historical customer demand spatiotemporal preference feature extraction vector, feature extraction processing is performed on the multiple historical customer demand information to obtain multiple historical customer demand spatiotemporal preference feature information. Based on the preset historical customer demand correlation feature extraction vector, feature extraction processing is performed on the multiple historical customer demand information to obtain multiple historical customer demand correlation feature information. The frequency characteristics of multiple historical customer demands are matched with the spatiotemporal preference characteristics of multiple historical customer demands to obtain spatiotemporal preference frequency matching information of multiple historical customer demands. The data is fused based on the interrelationship of multiple historical customer demand features and the spatiotemporal preference frequency matching information of multiple historical customer demand to generate multiple historical customer demand feature information.

4. The customer demand forecasting method as described in claim 1, characterized in that, The step of classifying and processing the multiple historical customer demand feature information to generate multiple historical customer demand group information specifically includes: Based on the preset historical customer demand group quantity information, the multiple historical customer demand feature information is randomly extracted to obtain multiple historical customer demand central feature information. Based on the aforementioned multiple historical customer demand characteristic information and multiple historical customer demand central characteristic information, multiple historical customer demand peripheral characteristic information are obtained; The Manhattan distance between the central feature information of the multiple historical customer demands and the peripheral feature information of the multiple historical customer demands is calculated to obtain the radial distance information of the multiple historical customer demand features; the radial distance information of the historical customer demand features corresponds one-to-one with the central feature information of the historical customer demands and the peripheral feature information of the historical customer demands. The minimum value of the radial distance information of the multiple historical customer demand features corresponding to the multiple historical customer demand peripheral feature information is used as the group confirmation distance information of the multiple historical customer demand peripheral feature information. Based on the confirmation distance information of the surrounding characteristic groups of the multiple historical customer needs and the central characteristic information of the historical customer needs, the surrounding characteristic information of the multiple historical customer needs is classified and processed to generate multiple historical customer needs to be confirmed group information. The information on the multiple historical customer demand groups to be confirmed is iteratively classified to obtain the surrounding feature information of multiple historical customer demand groups.

5. The customer demand forecasting method as described in claim 4, characterized in that, The step of iteratively classifying the information of the multiple historical customer demand groups to obtain the surrounding feature information of multiple historical customer demand groups specifically includes: Calculate the mean of the information of the multiple historical customer demand groups to be confirmed, and obtain the central information of the multiple historical customer demand groups to be confirmed. Determine whether the central information of the multiple historical customer demand groups to be confirmed is the same as the central feature information of multiple historical customer demand groups; If so, then based on the information of the multiple historical customer demand groups to be confirmed, multiple historical customer demand group information will be generated; If not, the central information of the multiple historical customer demand groups to be confirmed is used as the central feature information of multiple historical customer demand groups, and the process returns to the step of obtaining the peripheral feature information of multiple historical customer demand groups based on the multiple historical customer demand feature information and the central feature information of multiple historical customer demand groups.

6. A customer demand forecasting system, characterized in that, include: The historical customer demand information acquisition module is used to acquire multiple historical customer demand information. The historical customer demand feature information generation module is used to extract and analyze the features of the multiple historical customer demand information to generate multiple historical customer demand feature information. The historical customer demand group information generation module is used to classify and process the multiple historical customer demand feature information to generate multiple historical customer demand group information. The current customer demand information generation module is used to generate multiple customer group demand prediction parameter group information based on the multiple randomly generated customer group demand prediction weight information and the multiple randomly generated customer group demand prediction bias information; the customer group demand prediction parameter group information includes customer group demand prediction weight information and customer group demand prediction bias information. Based on the information of multiple historical customer demand groups and the information of multiple customer group demand prediction parameter groups, calculate the demand prediction accuracy representation information of multiple customer groups. Determine whether the maximum value of the multiple customer group demand prediction accuracy representation information is greater than or equal to the preset customer group demand prediction accuracy representation threshold. If so, then based on the customer group demand prediction parameter group information corresponding to the maximum value of the multiple customer group demand prediction accuracy characterization information and multiple historical customer demand group information, prediction calculation is performed to obtain multiple current customer demand information. If not, the customer group demand prediction parameter group information corresponding to the maximum value of the multiple customer group demand prediction accuracy characterization information shall be used as the core parameter group information for customer group demand prediction. Based on the information of the multiple customer group demand prediction parameter groups and the information of the core customer group demand prediction parameter groups, we obtain information of the multiple customer group demand prediction peripheral parameter groups. Based on the core parameter group information for customer group demand prediction, the preset customer group demand prediction weight adjustment step size, and the preset customer group demand prediction bias adjustment step size, the peripheral parameter group information for multiple customer group demand predictions is optimized to obtain multiple peripheral parameter group information for customer group demand predictions to be optimized. Based on the core parameter group information for customer group demand prediction and the peripheral parameter group information for customer group demand prediction to be optimized, multiple parameter groups for customer group demand prediction to be optimized are generated. The information of the multiple customer group demand prediction parameter groups to be optimized is used as the multiple customer group demand prediction parameter group information, and the process is returned to the step of calculating the multiple customer group demand prediction accuracy characterization information based on the multiple historical customer demand group information and the multiple customer group demand prediction parameter group information.

7. A terminal device, characterized in that, The terminal device includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 5.

Images