A service system based on automation technology to assist in completing the tender document production

Abstract

The application discloses a kind of service systems based on automation technology auxiliary completion tender document production, belong to tender service technical field, specifically include: user behavior acquisition module collects the operation behavior data of user in the process of tender document production;Decision preference analysis module generates user decision preference parameters by processing operation behavior data through analysis model;Strategy generation module generates candidate tender strategy set based on historical cases;Scene simulation module constructs tender scene simulation environment and records market competition data in the process of strategy execution;Strategy evaluation module calculates strategy competitiveness index according to market competition data;Strategy execution module filters optimal tender strategy and is converted into specific tender document production scheme;The application effectively improves the accuracy and competitiveness of tender document production by intelligently analyzing user behavior and dynamically simulating tender environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of bidding service, in particular to a service system for assisting in completing bidding document production based on automation technology. BACKGROUND

[0002] With the increasingly fierce market competition, bidding activities have become an important way for enterprises to obtain commercial projects. As a key carrier for showing the comprehensive strength of an enterprise, the quality of the bidding document directly affects the bidding result. Currently, various organizations need to invest a lot of manpower and resources in bidding document preparation during the bidding process, which involves multiple complex links such as market analysis, strategy formulation, and scheme design. Under the background of digital transformation, how to improve the efficiency and quality of bidding document production through technical means has become a widespread concern in the industry.

[0003] In the prior art, various bidding assistance systems have appeared. These systems usually provide a basic bidding document framework for users through rule matching based on a document template library and a historical case library. Some systems can automatically fill in the corresponding content paragraphs according to the user input project basic information; some other systems use keyword extraction technology to recommend similar content modules from historical successful cases. These technical solutions to some extent reduce the user's clerical work burden and provide a standardized document generation path.

[0004] However, the existing technical solutions generally generate bidding strategies based on static rules and fixed templates, lacking the ability to deeply mine the potential needs of users. They cannot effectively capture the decision-making preferences and risk tendencies of users during the document editing process, resulting in a deviation between the generated bidding strategies and the real intentions of users. At the same time, they lack a dynamic simulation mechanism for the market competition environment, making it difficult to assess the adaptability and competitiveness of the strategies in the actual bidding environment. This disconnection between strategy generation and actual application directly affects the quality of the bidding document and the success rate of bidding. SUMMARY

[0005] The purpose of the present application is to provide a service system for assisting in completing bidding document production based on automation technology, which solves the following technical problems:

[0006] The existing bidding document generation system cannot accurately capture the potential needs of users during the editing process, resulting in a deviation between the generated bidding strategies and the real intentions of users; at the same time, it lacks the ability to dynamically simulate the market competition environment.

[0007] The purpose of the present application can be achieved by the following technical solutions:

[0008] A service system for assisting in completing bidding document production based on automation technology, comprising:

[0009] A user behavior collection module is configured to collect operation behavior data of the user in the process of making the bid document, and the operation behavior data includes content adjustment records, format modification frequency and chapter access time length.

[0010] A decision preference analysis module is configured to process the operation behavior data through a decision preference analysis model to generate user decision preference parameters, and the user decision preference parameters include a scheme selection tendency and a risk control preference.

[0011] A strategy generation module is configured to input the user decision preference parameters into a bid strategy generation model, and the bid strategy generation model generates a candidate bid strategy set based on historical bid cases.

[0012] A scenario simulation module is configured to construct a bid scenario simulation environment, execute each strategy in the candidate bid strategy set in the bid scenario simulation environment, and record market competition data in the process of strategy execution.

[0013] A strategy evaluation module is configured to calculate a strategy competitiveness index based on the market competition data in the process of strategy execution, and the strategy competitiveness index includes a scheme advantage degree and a resource utilization efficiency.

[0014] A strategy execution module is configured to filter an optimal bid strategy according to the strategy competitiveness index, convert the optimal bid strategy into a specific bid document making scheme and execute the specific bid document making scheme.

[0015] As a further scheme of the application, in the decision preference analysis module, the specific process of processing the operation behavior data through the decision preference analysis model is as follows:

[0016] Content deletion position data of the user in the process of editing the bid document is collected, deletion operation distribution density of each chapter segment is counted, a time distribution map of modification operation is established in combination with time sequence information in the text modification frequency data, and a user focus concentration area is identified by analyzing duration characteristics in the paragraph stay time length data.

[0017] The deletion operation distribution density, the modification operation time distribution map and the user focus concentration area data are input into a bid decision analysis network, the bid decision analysis network includes a feature extraction layer, a time sequence processing layer and a decision output layer, the feature extraction layer calculates spatial distribution characteristics, the time sequence processing layer analyzes time sequence patterns, and the decision output layer generates user decision preference parameters, and the user decision preference parameters include a scheme selection tendency parameter and a risk control preference parameter.

[0018] As a further scheme of the application, in S3, the specific process of inputting the user decision preference parameters into the bid strategy generation model is as follows:

[0019] input the user decision preference parameter sequence into the bidding strategy encoder network, the bidding strategy encoder network adopts a commercial decision long short-term memory architecture to process sequence data, respectively calculates a forward commercial feature sequence and a backward commercial feature sequence, and fuses the forward commercial feature sequence and the backward commercial feature sequence to generate a complete commercial feature vector; the complete commercial feature vector is transmitted to a bidding strategy decoder network, the bidding strategy decoder network applies a commercial attention mechanism to calculate commercial context vectors of different dimensions, and integrates the commercial context vectors of different dimensions into a comprehensive commercial context vector;

[0020] a bidding strategy description sequence is gradually generated based on the comprehensive commercial context vector, each strategy description unit corresponds to a specific commercial strategy element; a commercial strategy search process is performed on the bidding strategy description sequence, a plurality of candidate strategy description sequences are maintained, and sequence generation probabilities are calculated in real time, the number of candidate strategies is dynamically adjusted according to the sequence generation probabilities, commercial strategy descriptions with repeated contents are removed, and an optimized candidate bidding strategy set is output.

[0021] As a further scheme of the application: the specific process in which the bidding strategy encoder network processes the user decision preference parameter sequence is:

[0022] the forward commercial state and the backward commercial state of the commercial decision long short-term memory network are initialized; each commercial parameter vector in the user decision preference parameter sequence is input in turn, and the forward commercial state sequence and the backward commercial state sequence are updated respectively; a final state vector of the forward commercial state sequence is calculated, and an initial state vector of the backward commercial state sequence is calculated;

[0023] the final state vector of the forward commercial state sequence and the initial state vector of the backward commercial state sequence are combined into a complete commercial feature vector; attention distribution data is calculated through a commercial attention mechanism to generate an enhanced commercial context vector; and the final commercial feature vector and the commercial context vector are output to the bidding strategy decoder network.

[0024] As a further scheme of the application: in the scene simulation module, the specific process of constructing the bidding scene simulation environment is:

[0025] the virtual bidding environment basic parameters are initialized, and the complete case records in the historical bidding project database are loaded; a plurality of competitor commercial models are configured, and a bidding strategy parameter set and a market reaction parameter set of each competitor commercial model are set, the bidding strategy parameter set includes an initial bid parameter, a bid adjustment parameter and a bid limit parameter, and the market reaction parameter set includes a reaction time parameter and a competition intensity parameter; a bidding market environment variable is defined, including a demand change parameter, a price sensitivity parameter and a market capacity parameter;

[0026] The candidate bidding strategy set is input into the bidding simulation environment in sequence, a multi-round bidding competition simulation process is executed, each round of simulation process includes a bidding stage, a bid evaluation stage and a result confirmation stage; the bidding market state data is updated in real time in each round of simulation process, the detailed bidding data, the bid-winning result data and the project income data of each bidding strategy are recorded, and the complete bidding simulation process data is stored to the bidding analysis database.

[0027] As a further scheme of the present application, the configuration process of the competitor business model is specifically:

[0028] The competitor basic attribute parameters are defined, the basic attribute parameters include competitor identifiers, historical bidding behavior data and business decision preference parameters; the competitor decision rule system is set, the decision rule system includes bid calculation rules, market reaction rules and risk assessment rules; the competitor instance object is initialized, and an initial business state parameter set is allocated; the decision rule system is called in real time in the bidding simulation process, and the bid decision data is generated according to the current bidding market state; the competitor internal state parameters are updated, and the complete competitor behavior log data is recorded, the competitor behavior log data includes decision time points, decision input information and decision output results.

[0029] As a further scheme of the present application, in the strategy evaluation module, the specific process of calculating the strategy competitiveness index based on the market competition data in the strategy execution process is:

[0030] The complete bid sequence data of each bidding strategy is extracted from the bidding analysis database, the business deviation degree of the bid sequence and the market average bid is calculated, the business deviation degree includes absolute difference data and relative fluctuation data; the market share time series data of each bidding strategy is analyzed, the instant growth data and long-term trend data of the market share are calculated; the scheme advantage degree index is generated by business weight calculation combined with the bid business deviation degree and the market share growth characteristics;

[0031] Meanwhile, the resource consumption time series data of each bidding strategy is extracted, the instant efficiency data and long-term efficiency data of resource input and output are calculated, and the resource utilization efficiency index is generated; the scheme advantage degree index and the resource utilization efficiency index are processed by business standardization, and the bidding strategy comprehensive competitiveness score is generated by business combination calculation.

[0032] As a further scheme of the present application, in the strategy execution module, the specific process of selecting the optimal bidding strategy according to the strategy competitiveness index is:

[0033] Receiving the bidding strategy comprehensive competitiveness score data, selecting the bidding strategy scheme with the highest score as the target bidding strategy; analyzing the complete content description of the target bidding strategy, identifying the bidding document paragraph position that needs to be adjusted and the content modification requirements; generating a detailed bidding document modification instruction set, the bidding document modification instruction set including modification position information, modification content text and modification format requirements; performing document modification operations according to the bidding document modification instruction set, completing the bidding document content update and format specification adjustment; using the bidding document processing engine to generate the final document, the bidding document processing engine outputting the standardized format bidding document according to the bidding document specification requirements.

[0034] Advantages of the present application:

[0035] The present application effectively solves the technical problem that the existing bidding technology cannot capture the potential needs of users by constructing a user behavior collection and decision preference analysis model. By analyzing the user's operation behaviors such as deletion position, modification frequency and residence time during document editing, a precise user preference portrait is established, and a bidding strategy that meets the user's real intention is generated. Further, by establishing a virtual bidding scene simulation environment, the system can simulate multiple rounds of market competition for candidate strategies, and dynamically evaluate the adaptability and competitiveness of the strategies combined with real-time market data. The multi-objective optimization screening mechanism based on strategy competitiveness indicators ensures that the final output bidding strategy has both scheme advantage and resource utilization efficiency. The entire scheme forms a complete technical closed loop from behavior collection, strategy generation, simulation evaluation to optimization output, and continuously optimizes the strategy generation effect through a continuous learning mechanism, significantly improving the quality and market competitiveness of the bidding document, and effectively improving the bidding success rate. BRIEF DESCRIPTION OF DRAWINGS

[0036] The present application will be further described below with reference to the accompanying drawings.

[0037] Figure 1 is a module schematic diagram of the present application. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0039] Please refer to Figure 1 As shown in the figure, the present application is a service system for assisting the completion of bidding document production based on automation technology, which comprises:

[0040] The user behavior collection module is configured to collect operation behavior data of the user in the process of making the bidding document, and the operation behavior data includes content adjustment records, format modification frequency, and chapter access time length.

[0041] The decision preference analysis module is configured to process the operation behavior data by using a decision preference analysis model to generate user decision preference parameters, and the user decision preference parameters include a scheme selection tendency and a risk control preference.

[0042] The strategy generation module is configured to input the user decision preference parameters into a bidding strategy generation model, and the bidding strategy generation model generates a candidate bidding strategy set based on historical bidding cases.

[0043] The scene simulation module is configured to construct a bidding scene simulation environment, execute each strategy in the candidate bidding strategy set in the bidding scene simulation environment, and record market competition data in the process of strategy execution.

[0044] The strategy evaluation module is configured to calculate a strategy competitiveness index based on the market competition data in the process of strategy execution, and the strategy competitiveness index includes a scheme advantage degree and a resource utilization efficiency.

[0045] The strategy execution module is configured to filter an optimal bidding strategy according to the strategy competitiveness index, convert the optimal bidding strategy into a specific bidding document making scheme, and execute the specific bidding document making scheme.

[0046] In a preferred embodiment of the present application, the specific process of processing the operation behavior data by using the decision preference analysis model in the decision preference analysis module is as follows:

[0047] First, an event listener is deployed in the bidding document editing interface, and the listener is configured to capture three types of operation behaviors of the user in the process of document editing in real time. For content deletion operation, the system records the specific position of each deletion, including paragraph number, sentence index and character offset, and records the type of deleted content, which is technical description, business terms or price information. For example, when the user frequently deletes specific terms in the technical scheme part, the system records that these deletion operations are concentrated in the 3rd paragraph of the document, and mainly involve "cloud computing architecture" related expressions.

[0048] The system is configured to count operation data in time windows, and each time window is set to 30 minutes. In a single time window, the system calculates the deletion operation density of each paragraph, and the specific method is to divide the number of deletions by the paragraph length, and then multiply by 100 to obtain a standardized density value. For example, the length of a technical scheme paragraph is 500 characters, and 5 deletion operations occur within 30 minutes, so the deletion density is calculated as 1.0.

[0049] For text modification frequency data, the system records the timestamp and modification type of each modification. Modification types are classified into three categories: content replacement, format adjustment, and structure adjustment. The system establishes a time distribution map of modification operations at 15-minute intervals, which contains 24 time points, and each time point records the number of modification operations in that period. For example, 8 content replacements and 3 format adjustments are detected in the 9:00-9:15 period, so the modification intensity at this time point is marked as 11.

[0050] When analyzing paragraph dwell time, the system identifies the user's reading pattern by monitoring cursor position changes. When the cursor stays in a certain paragraph area for more than 10 seconds, the system starts recording the dwell time. At the same time, the system identifies abnormal dwell situations, such as a certain budget explanation paragraph with an average dwell time of 45 seconds, while other paragraphs have an average dwell time of only 15 seconds, which is marked as a focus concentration area.

[0051] These preliminary processed behavior data are then input into the bidding decision analysis network. The network adopts a three-layer architecture design, the first layer is the feature extraction layer, which uses 128 convolution kernels to process spatial distribution features, with a convolution kernel size of 3x3 and a step size of 1. This layer is specifically designed to process deletion operation distribution density data, and identifies the user's focus area pattern through convolution operations.

[0052] The second layer is the time series processing layer, which uses a long short-term memory network structure. This layer contains 64 memory units and processes the time distribution map data of modification operations. This network layer captures the time regularity of user operations through memory gating mechanisms, such as identifying that users tend to modify technical content in the morning and focus on business clause adjustments in the afternoon.

[0053] The third layer is the decision output layer, which fuses the features extracted by the previous two layers and generates user decision preference parameters through a fully connected neural network. The fully connected layer contains 3 hidden layers with 256, 128, and 64 nodes respectively, using ReLU activation functions. The final output layer generates two dimensions of decision parameters: scheme selection preference parameters and risk control preference parameters.

[0054] In specific implementation, the scheme selection preference parameter is calculated by analyzing the user's different processing methods for technical schemes and business schemes. For example, when the user frequently modifies the depth and breadth of the technical scheme, but rarely adjusts the business terms, the system will judge that the user is more concerned about the selection of the technical scheme. The risk control preference parameter is inferred by analyzing the user's editing frequency of risk clauses, disclaimers, etc. For example, if the user repeatedly modifies the content of risk response measures, the system will judge that the user has a high risk control demand.

[0055] During the whole process, the system updates the data using a sliding window mechanism, recalculates the user decision preference parameters every 5 minutes to ensure that it can capture the changes in user demand in a timely manner. All intermediate processing results are saved in the memory database for subsequent module calling analysis.

[0056] In another preferred embodiment of the application, in S3, the specific process of inputting the user decision preference parameters into the bidding strategy generation model is as follows:

[0057] The bidding strategy encoder network adopts a bidirectional long short-term memory architecture, which is composed of two independent LSTM networks, respectively responsible for forward and backward sequence processing. In the network initialization stage, the hidden state and cell state of the forward business state are set to zero vectors, and the vector dimension is 256; similarly, the hidden state and cell state of the backward business state are also initialized to zero vectors of the same dimension.

[0058] The user decision preference parameter sequence is input into the encoder network in chronological order. Each business parameter vector contains 16-dimensional feature data, including 8-dimensional scheme selection inclination parameters and 8-dimensional risk control preference parameters. The sequence length is fixed at 32 time steps, corresponding to the 32 user decision preference snapshots recorded by the system recently. For each input vector, the forward LSTM network processes from the 1st time step to the 32nd time step in chronological order, and calculates the hidden state of each time step. At the same time, the backward LSTM network processes from the 32nd time step to the 1st time step in reverse order, and independently calculates its hidden state sequence.

[0059] At the time of processing the 1st time step, the forward LSTM receives the initial hidden state and the first business parameter vector, and updates the cell state and outputs the first forward hidden state through the calculation of the input gate, the forgetting gate and the output gate. The forgetting gate uses the sigmoid activation function to calculate the proportion of historical information retained, the input gate determines the degree of new information added, and the output gate controls the output content of the current hidden state. The calculation of these gate mechanisms involves weight matrices and bias terms, and the dimension of the weight matrix is 256 by 272, where 272 is the splicing dimension of the 256-dimensional hidden state and the 16-dimensional input vector.

[0060] As the sequence processing progresses, the forward LSTM gradually builds a hidden state sequence containing historical information of the sequence. For example, when processing to the 16th time step, the forward hidden state has fused the user preference features of the first 16 time steps. At the same time, the backward LSTM starts processing from the end of the sequence, and when it processes to the 16th time step, its hidden state already contains the feature information of the 17th to 32nd time steps.

[0061] When the sequence processing is completed, the forward LSTM outputs the hidden state vector of the final time step, which has a dimension of 256 and contains the forward context information of the entire sequence. The backward LSTM outputs the hidden state vector of its initial time step, also with a dimension of 256, containing the backward context information of the entire sequence. These two vectors are merged into a complete business feature vector of 512 dimensions through a concatenation operation.

[0062] The complete business feature vector is then passed to the bid strategy decoder network. The decoder network adopts an attention mechanism-based sequence-to-sequence architecture, with its initial hidden state set to the complete business feature vector output by the encoder. The decoder processes all the hidden states of the encoder through a multi-head attention mechanism, including the 32 hidden states of the forward LSTM and the 32 hidden states of the backward LSTM, a total of 64 hidden states.

[0063] The attention mechanism calculation is divided into 8 attention heads, each responsible for a different semantic space. For each attention head, the query vector, key vector, and value vector are first calculated. The query vector comes from the current hidden state of the decoder, and the key vector and value vector come from the hidden state set of the encoder. Each attention head independently calculates the attention weight distribution, which is normalized by the softmax function to obtain the attention distribution of the 64 hidden states.

[0064] For example, the first attention head may pay more attention to the hidden states related to the technical solution, with higher values in the technical feature dimension; the second attention head may pay more attention to the hidden states related to risk control, allocating more weights in the risk feature dimension. Each attention head generates a corresponding context vector by weighting and summing the value vector according to its attention weights.

[0065] The context vectors generated by the 8 attention heads are merged into a comprehensive business context vector of 2048 dimensions through a concatenation operation, which is reduced to 512 dimensions by a linear layer and fused with the current hidden state of the decoder. The fused vector is processed by a tanh activation function and input to the output layer.

[0066] The output layer is a fully connected neural network that maps the 512-dimensional input vector to a probability distribution of 1024 strategy tokens. The strategy token vocabulary contains common business strategy elements, such as differentiated pricing, emphasis on technical advantages, risk response plans, etc. Each strategy token corresponds to a specific strategy description unit, such as using a step pricing strategy to highlight technical innovation capabilities and emphasizing after-sales service commitments, etc.

[0067] The decoder generates the strategy description sequence in an autoregressive manner, starting from the start token and generating each strategy token step by step. When generating each new token, the decoder updates its hidden state and recalculates the attention weights. The generation process continues until the end token is produced or the maximum sequence length of 50 tokens is reached.

[0068] Throughout the generation process, the system performs a beam search algorithm with a beam width of 5. This means that the system maintains 5 candidate sequences at any time, each with a corresponding cumulative log probability. When a new token is generated, each candidate sequence is expanded into the top 5 new sequences in terms of probability, and then the top 5 sequences in terms of cumulative probability are selected from the 25 candidate sequences to continue generation.

[0069] For example, when generating the 10th token, a certain candidate sequence may already contain 9 strategy tokens such as technical solution innovation, emphasis on cost control, and highlighting of localized services, with a cumulative log probability of -12.5. The system calculates the probability distribution of the next token based on the current hidden state and attention context, and selects the top candidate tokens with the highest probability for expansion.

[0070] The system monitors the generation quality of candidate sequences in real time and optimizes the output results through the following mechanisms: first, the system calculates the repetition penalty of each candidate sequence, which reduces the probability weight of repeated strategy tokens; second, the system implements length normalization, dividing the cumulative probability by the alpha power of the sequence length, where alpha is set to 0.8 to balance the influence of sequence length; finally, the system uses the coverage mechanism to prevent the attention from repeatedly focusing on the same encoder state.

[0071] When all candidate sequences generate the end token or reach the maximum length, the system selects the sequence with the highest cumulative probability as the main output. At the same time, the system retains the top 5 best sequences as alternative strategies. These strategy description sequences go through a post-processing module to remove repetitive semantic content, ensuring that each strategy description unit has a unique value proposition.

[0072] For example, a generated strategy description sequence may contain the following sequence of strategy tokens: adopt a differentiated pricing strategy, set the benchmark price at 95% of the market level, but use a premium pricing strategy for core technology services with a premium of 15%, emphasize the advantages of the localized service team, provide a 24-hour response commitment, and propose a three-level security mechanism in risk control.

[0073] The final output candidate bidding strategy set contains 3 to 5 complete strategy descriptions, each of which contains core elements such as bid strategy, technical solution focus, service commitment, and risk control measures. These strategy descriptions are ranked according to the comprehensive score, which takes into account the sequence generation probability, strategy innovation, and matching degree with user preferences.

[0074] In another preferred embodiment of the present application, in the scenario simulation module, the specific process of constructing the bidding scenario simulation environment is:

[0075] First, load complete case records from the historical bidding project database for the past three years, which cover bidding projects in different industry fields, including information technology services, engineering construction, equipment procurement, etc. The system loads 1000 project cases, each containing complete bidding documents, evaluation criteria, competitor bids, bid results, and subsequent execution data. The environment base parameter settings include a simulation time span of 90 days, a simulation granularity of one day, and a market area divided into five major economic zones.

[0076] The construction of the competitor business model uses an agent-based modeling method. The system creates an independent business model instance for each competitor, each instance containing complete attribute parameters and decision rules. Competitor basic attribute parameters include competitor identifiers, represented by an 8-digit code, with the first two digits representing industry categories, the middle three digits representing enterprise size, and the last three digits representing regional distribution. Historical bidding behavior data is extracted from the database, including the number of times the competitor has participated in bidding, the bid rate, typical bid strategies, and technical advantage areas. Business decision preference parameters are derived by analyzing historical bidding data, including risk tolerance, innovation inclination, and price sensitivity, each dimension quantified as a value between 0 and 100.

[0077] The competitor decision rule system uses a combination of rule engines and machine learning. The bid calculation rules include a basic bid calculation module and a dynamic adjustment module. The basic bid calculation module generates an initial bid using a random forest algorithm based on project size, complexity, and historical bid data. For example, for a 500,000 yuan information technology project, competitor A's initial bid may be calculated as 85% of the budget, i.e. 425,000 yuan. The bid adjustment module dynamically adjusts the bid based on market feedback during the simulation process, with the adjustment range influenced by factors such as current production capacity utilization, market share target, and cash flow status.

[0078] The market response rule is implemented through a deep Q-learning algorithm, which is trained based on historical behavior data of competitors. Each competitor's business model includes an independent neural network, with 15-dimensional market state features such as current market share, recent bid success, and other competitor bids in the input layer, and 8 possible action choices such as maintaining the bid, reducing the bid by 3%, reducing the bid by 5%, and increasing the bid by 2% in the output layer. The network hidden layer contains 3 layers with 128 nodes each, using ReLU activation function.

[0079] The risk assessment rule uses a multi-factor weighted evaluation model, considering three dimensions of project risk, competition risk, and execution risk. Project risk includes factors such as technical difficulty, project duration, and payment conditions; competition risk includes factors such as the number of competitors and the strength of competitors; and execution risk includes factors such as resource matching and similar project experience. Each factor is assigned a different weight based on historical data, and a comprehensive risk coefficient is finally calculated.

[0080] When initializing the competitor instance object, the system assigns an initial set of business state parameters to each competitor, including current market share, financial status, technical ability score, and brand influence. Market share is set based on market research data in the last quarter, financial status is derived by analyzing public financial statements and industry research data, technical ability score is based on the number of patents, proportion of technical personnel, and past project quality assessment, and brand influence is calculated by market research and media exposure.

[0081] The execution process of the bidding simulation environment is divided into multiple rounds, with each round representing a complete bidding cycle. The system first initializes market environment variables, including demand change parameters, price sensitivity parameters, and market capacity parameters. Demand change parameters are predicted through time series analysis, considering seasonal fluctuations, industry development cycles, and policy impacts. Price sensitivity parameters are calculated through market research data, with different price elasticity coefficients set for different industries. Market capacity parameters are determined based on macroeconomic data and industry reports.

[0082] After inputting the candidate bidding strategy set into the simulation environment, the system begins to execute multiple rounds of bidding competition simulation. Each simulation round includes three distinct stages: bidding stage, bid evaluation stage, and result confirmation stage. In the bidding stage, all competitor business models generate bidding schemes in parallel based on the current market state, including technical schemes, commercial schemes, and bid schemes. Technical schemes highlight their own strengths, commercial schemes emphasize differentiated service commitments, and bid schemes are calculated based on their respective decision-making rules.

[0083] The bid evaluation stage simulates the bid evaluation process of a real project. The system quantitatively scores each bid scheme according to the pre-set bid evaluation criteria. The bid evaluation criteria include technical scheme score, business scheme score and price score, and the weight proportion is set according to the project type. For example, for a technical innovation project, the technical scheme weight may be set to 50%, the business scheme weight to 30%, and the price weight to 20%. The technical scheme score is calculated by analyzing the innovativeness, feasibility and integrity of the scheme content through natural language processing technology; the business scheme score is calculated by evaluating after-sales service, delivery capacity and risk control measures; and the price score is calculated by using the benchmark price method, taking the average value of all valid bids as the benchmark.

[0084] The result confirmation stage determines the winning bidder according to the bid score and updates the business status of all competitors. The market share of the winning bidder is correspondingly improved, the financial condition is improved, and the brand influence is enhanced. The non-winning bidders adjust their business strategies according to the bidding results, such as strengthening technical research and development investment, adjusting bidding strategies or exploring new market fields.

[0085] During the entire simulation process, the system updates the bid market state data in real time, records the detailed bidding data, bid result data and project income data of each bidding strategy in each round. The bidding data includes initial bid, final bid and bid adjustment history; the bid result data includes score details, ranking and elimination reason analysis; and the project income data includes direct profit, brand value improvement and potential business opportunity evaluation.

[0086] The simulation process data is completely stored in the bid analysis database, and the database uses a time series data model, each data point containing three fields of timestamp, data type and data value. The system generates an independent session identifier for each simulation run to ensure the integrity and traceability of the data. For example, a complete simulation run may include 30 rounds, generating about 15000 data records, covering the behavior decisions and market reactions of all competitors in each round.

[0087] The competitor behavior log data is recorded in a structured storage manner, and each decision time point records the decision input information and decision output result. The decision input information includes the current market state snapshot, the self business state and the competitor dynamics; the decision output result includes the specific action taken, the expected income evaluation and the risk evaluation conclusion. These log data are used for subsequent analysis and model optimization, helping to understand the behavior patterns and decision logic of competitors.

[0088] In another preferred embodiment of the present application, in the strategy evaluation module, the specific process of calculating the strategy competitiveness index based on the market competition data in the strategy execution process is:

[0089] Firstly, the system extracts the complete dataset generated in the simulation process from the bidding analysis database. These data are organized according to the simulation rounds, and each round contains detailed execution records of each bidding strategy. The system extracts the bid sequence data of all bidding strategies in a 30-day analysis period. The bid sequence data includes daily bid records, bid adjustment time points, and adjustment amplitudes, etc.

[0090] When the system calculates the business deviation degree of the bid sequence and the market average bid, a double-index evaluation system is adopted. The absolute difference data is obtained by calculating the difference between the strategy bid and the market average price at each time point. The difference value is divided into positive deviation and negative deviation. For example, if a strategy bid is 4.5 million yuan at a certain time point, and the market average bid is 4.8 million yuan, the absolute difference is -300,000 yuan. The relative volatility data is obtained by calculating the coefficient of variation of the bid sequence. First, the standard deviation of each strategy bid sequence is calculated, and then divided by the average value of the sequence to obtain the relative volatility rate. For example, if the standard deviation of a strategy bid in 30 days is 250,000 yuan, and the average bid is 460,000 yuan, the relative volatility rate is 5.43%.

[0091] When analyzing the market share time series data, the system uses a sliding window technique to process the data. The instantaneous growth data is obtained by calculating the market share change rate between adjacent time points. Specifically, it calculates the difference between today's market share and yesterday's market share, and then divides it by yesterday's market share. For example, if a strategy's market share grows from 15% yesterday to 18% today, the instantaneous growth rate is 20%. The long-term trend data is obtained by linear regression analysis. The system fits a trend line to the 30-day market share data and calculates the slope value of the trend line. A positive slope indicates an upward trend in market share, and a negative slope indicates a downward trend.

[0092] The calculation of the scheme advantage degree index adopts a multi-factor weighted evaluation model. This model includes two first-level indicators: business deviation degree and market share growth characteristics. The business deviation degree index is divided into absolute difference sub-item and relative volatility sub-item; the market share growth characteristics index is divided into instantaneous growth sub-item and long-term trend sub-item. Each sub-item is standardized to convert it to a score value of 0 to 100. The weight distribution is determined according to expert evaluation method and historical data statistical analysis. The total weight of business deviation degree is 45%, of which the absolute difference weight is 25% and the relative volatility weight is 20%; the total weight of market share growth characteristics is 55%, of which the instantaneous growth weight is 30% and the long-term trend weight is 25%.

[0093] The extraction of resource consumption time series data covers three dimensions of manpower investment, capital investment and time investment. The manpower investment data includes the number of personnel involved in the project and the work duration record; the capital investment data includes direct cost expenditure and indirect cost allocation; the time investment data includes the time for scheme preparation, bid response and subsequent service. The system aggregates these data on a weekly basis to form the time series of resource consumption.

[0094] The instant efficiency data is obtained by calculating the input-output ratio within a single week. The output indicators include the amount of bid won, the number of new customers acquired and the brand exposure, etc. For example, in a certain week, the input manpower resources are 200 man-hours, the capital cost is 150,000 yuan, and the amount of bid won is 1,000,000 yuan. Then the instant efficiency is calculated as the output divided by the input, which is 3.33. The long-term efficiency data is obtained by calculating the cumulative input-output ratio within the entire analysis period, taking into account the time value of money, and using the net present value method to standardize the input and output at different time points.

[0095] The calculation of resource utilization efficiency indicators uses the analytic hierarchy process to establish a judgment matrix for pairwise comparison of the three resource dimensions to determine the relative importance of each dimension. After consistency check, the weight of manpower investment is 35%, the weight of capital investment is 40%, and the weight of time investment is 25%. The efficiency value of each dimension is standardized, and the resource utilization efficiency indicator is obtained by weighted summation.

[0096] The business standardization process uses the min-max normalization method to map the original values of the scheme advantage degree indicator and the resource utilization efficiency indicator to the standard score range of 0 to 100. The specific method is to first determine the minimum and maximum values of each indicator in the historical data set, and then calculate each data point using the formula. For example, the historical minimum value of the scheme advantage degree indicator is 20 points, the maximum value is 90 points, and the original value of a certain strategy is 75 points. The standardized score is 79.3 points.

[0097] The bid strategy comprehensive competitiveness score is calculated by linear weighted combination. The weight of the scheme advantage degree indicator is 60%, and the weight of the resource utilization efficiency indicator is 40%. The weight setting is based on the regression analysis result of the historical bid data to ensure that the score result has high correlation with the real bid rate. Each strategy finally gets a comprehensive score of 0 to 100, and the higher the score, the stronger the competitiveness.

[0098] In another preferred embodiment of the present application, in the strategy execution module, the specific process of selecting the optimal bid strategy according to the strategy competitiveness indicator is as follows:

[0099] After receiving this comprehensive competitiveness score data, the strategy execution module first sorts the scores from highest to lowest and selects the bidding strategy with the highest score as the target bidding strategy. For example, in an evaluation, if strategy A scores 92 points, strategy B scores 87 points, and strategy C scores 83 points, then strategy A is selected as the target strategy.

[0100] When parsing the complete description of the target bidding strategy, the system uses natural language processing technology to identify key elements in the strategy text. These elements include the pricing strategy description, key technical solutions, service commitment terms, and risk control measures. The system establishes a mapping relationship between these elements and paragraphs in the bid document; for example, the pricing strategy description corresponds to the commercial pricing section of the bid document, and the key technical solutions section corresponds to the technical solutions section.

[0101] The modification location identification process is based on the standard template structure of the tender documents. The system maintains an index containing all standard chapters and paragraphs. Each paragraph has a unique identifier; for example, the second paragraph of section three in the technical solution chapter is identified as TECH3.2. The system determines the specific location that needs to be modified based on the correspondence between strategy elements and paragraph identifiers.

[0102] The set of bid document modification instructions is generated using a structured representation method. Each modification instruction contains four fields: operation type, target location, modification content, and formatting requirements. Operation types include insertion, deletion, and replacement; the target location uses paragraph identifiers for precise positioning; the modification content provides the specific text content; and the formatting requirements specify layout parameters such as font, font size, and spacing. For example, a specific modification instruction might be placed in the fifth paragraph of the commercial quotation section, replacing the original quotation description with a new tiered pricing scheme, and using SimSun font, size 12 (small four).

[0103] Document modification operations are executed using a transaction processing mechanism, ensuring that all modifications either succeed completely or are rolled back completely. The system first creates a copy of the tender document, and then executes each modification command sequentially on the copy. During execution, syntax checks and format validations are performed to ensure that the modified document meets the specifications. For example, replacement operations check whether the length of the new text is appropriate for the original position, and format adjustments verify whether the font and font size meet the standards.

[0104] The tender document processing engine employs template-based document generation technology. The engine incorporates multiple standard tender document templates covering various industries and project types. Based on the modified document content and formatting requirements, the engine selects the appropriate template for rendering. The rendering process includes steps such as text flow layout, chart insertion, and page number generation, ultimately outputting a tender document that conforms to industry standards.

[0105] The standardized format output ensures that the bidding document meets the technical requirements of the tendering party, including file format, font specification, page layout, and directory structure, etc. The system supports output of bidding documents in multiple formats, including PDF format, Word format, and HTML format. Before output, final quality checks are performed, including spelling checks, format consistency, and content integrity verification.

[0106] The above has carried out the detailed explanation to one embodiment of the application, but the content described is only the preferred embodiment of the application, cannot be considered for limiting the implementation range of the application. All equivalent changes and improvements made in the scope of the application are still within the scope of the patent coverage of the application.

Claims

1. A service system for assisting completion of a bid document based on automation technology, characterized by, The method comprises the following steps: a user behavior collection module is used to collect the operation behavior data of the user in the process of making the bidding document, and the operation behavior data comprises content adjustment records, format modification frequency and chapter access time length; a decision preference analysis module is used to process the operation behavior data through a decision preference analysis model to generate user decision preference parameters; a strategy generation module is used to input the user decision preference parameters into a bidding strategy generation model, and the bidding strategy generation model generates a candidate bidding strategy set based on historical bidding cases, which comprises the following steps: the user decision preference parameter sequence is input into a bidding strategy encoder network, the bidding strategy encoder network adopts a business decision long short-term memory architecture, and forward business feature sequences and backward business feature sequences are calculated respectively, the forward business feature sequences and the backward business feature sequences are fused to generate a complete business feature vector; the complete business feature vector is transmitted to a bidding strategy decoder network, and the bidding strategy decoder network applies a business attention mechanism to calculate business context vectors of different dimensions, and integrates the business context vectors of different dimensions into a comprehensive business context vector; the bidding strategy description sequence is generated step by step based on the comprehensive business context vector, and each strategy description unit corresponds to a business strategy element; a business strategy search process is performed on the bidding strategy description sequence, a plurality of candidate strategy description sequences are maintained, and sequence generation probabilities are calculated in real time, the number of candidate strategies is dynamically adjusted according to the sequence generation probabilities, the business strategy descriptions with repeated contents are removed, and an optimized candidate bidding strategy set is output; a scene simulation module is used to build a bidding scene simulation environment, and each strategy in the candidate bidding strategy set is executed in the bidding scene simulation environment, and market competition data in the strategy execution process is recorded; a strategy evaluation module is used to calculate a strategy competitiveness index based on the market competition data in the strategy execution process; a strategy execution module is used to select an optimal bidding strategy according to the strategy competitiveness index, convert the optimal bidding strategy into a specific bidding document making scheme and execute the specific bidding document making scheme, which comprises the following steps: the bidding strategy scheme with the highest score is selected as a target bidding strategy; the complete content description of the target bidding strategy is analyzed, the bidding document paragraph position to be adjusted and the content modification requirement are identified; a bidding document modification instruction set is generated, the bidding document modification instruction set comprises modification position information, modification content text and modification format requirement; document modification operations are performed according to the bidding document modification instruction set, and content update and format specification adjustment of the bidding document are completed; a bidding document processing engine is used to generate a final document, and the bidding document processing engine outputs a standardized format bidding document according to the bidding document specification requirement.

2. The service system for assisting in making a bid document based on automation technology according to claim 1, characterized in that, In the decision preference analysis module, the specific process of processing the operation behavior data through the decision preference analysis model comprises the following steps: content deletion position data of the user in the process of editing the bidding document is collected, and deletion operation distribution density of each chapter paragraph is counted; a time distribution atlas of modification operations is established in combination with time sequence information in the text modification frequency data; the duration characteristics in the paragraph stay time length data are analyzed, and the user focus concentration area is identified. The deletion operation distribution density, the modification operation time distribution map and the user attention degree concentrated area data are input into a bid decision analysis network, the bid decision analysis network comprises a feature extraction layer, a time sequence processing layer and a decision output layer, the feature extraction layer calculates spatial distribution features, the time sequence processing layer analyzes time sequence modes, and the decision output layer generates user decision preference parameters, the user decision preference parameters comprise a scheme selection tendency parameter and a risk control preference parameter.

3. The service system for assisting in making a bid document based on automation technology according to claim 1, characterized in that, In the scene simulation module, the specific process of constructing the bidding scene simulation environment is: Initialize the virtual bidding environment basic parameters, load the complete case records in the historical bidding project database; configure multiple competitor business models, set the bidding strategy parameter set and the market reaction parameter set of each competitor business model, the bidding strategy parameter set includes the initial bid parameter, the bid adjustment parameter and the bid limit parameter, the market reaction parameter set includes the reaction time parameter and the competition intensity parameter; define the bidding market environment variables, including the demand change parameter, the price sensitivity parameter and the market capacity parameter; The candidate bidding strategy set is input into the bidding simulation environment in turn, and a plurality of rounds of bidding competition simulation processes are executed, each round of simulation process includes a bidding stage, an evaluation stage and a result confirmation stage; the bidding market state data is updated in real time in each round of simulation process, the detailed bid data, the bid result data and the project income data of each bidding strategy are recorded, and the complete bidding simulation process data is stored to the bidding analysis database.

4. The service system for assisting in making a bid document based on automation technology according to claim 3, characterized in that, The specific process of configuring the competitor business model is: Define the competitor basic attribute parameters, including the competitor identifier, the historical bidding behavior data and the business decision preference parameters; Set the competitor decision rule system, the decision rule system includes the bid calculation rule, the market reaction rule and the risk assessment rule; initialize the competitor instance object, and allocate the initial business state parameter set; In the bidding simulation process, the decision rule system is called in real time, and the bid decision data is generated according to the current bidding market state; Update the competitor internal state parameters, and record the complete competitor behavior log data, including the decision time point, the decision input information and the decision output result.

5. The service system for assisting in making a bid document based on automation technology according to claim 1, characterized in that, In the strategy evaluation module, the specific process of calculating the strategy competitiveness index based on the market competition data in the strategy execution process is: Extract the complete bid sequence data of each bidding strategy from the bidding analysis database, calculate the business deviation degree of the bid sequence and the market average bid, and the business deviation degree includes the absolute difference data and the relative fluctuation data; analyze the market share time sequence data of each bidding strategy, calculate the instant growth data and the long-term trend data of the market share; Combine the bid business deviation degree and the market share growth characteristics, and generate the scheme advantage degree index through business weight calculation; At the same time, the resource consumption time sequence data of each bidding strategy is extracted, the instant efficiency data and the long-term efficiency data of resource input and output are calculated, and the resource utilization efficiency index is generated; the scheme advantage degree index and the resource utilization efficiency index are subjected to business standardization processing, and the bidding strategy comprehensive competitiveness score is generated through business combination calculation.

Images