Big Data-based Intelligent Document and Management Collaborative Management System

Abstract

The present invention discloses a smart cultural and administrative collaborative management system based on big data, which specifically relates to the field of information technology. It includes a cultural data collection module, an analysis model construction module, an automatic monitoring module, and a collaborative response module. The cultural data collection module collects and preprocesses cultural and tourism contract data. The analysis model construction module uses data analysis technology to construct a cultural and tourism mathematical model and sets a reasonable price threshold for the cultural and tourism mathematical model. The automatic monitoring module conducts real-time monitoring on newly signed cultural and tourism contracts, uses machine learning algorithms to construct an abnormal contract detection model, identifies tourism contracts with abnormal prices, and classifies warning signals according to the degree of unreasonable pricing. The collaborative response module collects the process data of warning events according to the warning signals and stores them in the cloud, and pushes the warning events and process data to the law enforcement department to conduct investigation and supervision on the warning events, effectively solving the problem before the warning is reported and improving the efficiency of law enforcement supervision in the cultural field.

Description

Technical Field

[0001] The present invention relates to the field of information technology. More specifically, the present invention is a smart cultural and tourism management collaborative management system based on big data. Background Art

[0002] When a cultural and tourism institution signs a contract with a tourist, the cultural and tourism institution may use vague and unclear contract terms, making it difficult for tourists to detect or hold the unreasonable parts in the price accountable. There may be a large number of additional terms and fees in the contract, resulting in a huge gap between the actual payment and the advertised price. Or the fees of a low-price tour group are split into multiple parts, covering up the actual price through various hidden charges. Unreasonably low-price tour groups have always been a stubborn problem in the cultural and tourism industry. Market supervision departments, as well as public security and network information departments, often only learn about the warning signals of unreasonably low-price tour groups after receiving reports, and it is difficult to conduct timely and effective collaborative supervision of risk events when the bad signals first appear.

[0003] To solve the above defects, a technical solution is proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide a smart cultural and tourism management collaborative management system based on big data to solve the problems in the above background art.

[0005] To achieve the above purpose, the present invention provides the following technical solution: A smart cultural and tourism management collaborative management system based on big data, including a cultural data collection module, an analysis model construction module, an automatic monitoring module, and a collaborative response module;

[0006] The cultural data collection module is used to collect cultural and tourism contract data, and preprocess the cultural and tourism contract data and then send it to the analysis model construction module;

[0007] The analysis model construction module is used to construct a cultural and tourism mathematical model by using data analysis technology and set a reasonable price threshold for the cultural and tourism mathematical model;

[0008] The automatic monitoring module is used to monitor newly signed cultural and tourism contracts in real time, construct an abnormal contract detection model by using machine learning algorithms, identify tourism contracts with abnormal prices. When it detects that the price of a cultural and tourism contract is lower than the reasonable price threshold set for the cultural and tourism mathematical model, it starts a warning signal and classifies the warning signal according to the degree of unreasonable pricing;

[0009] The collaborative response module is used to collect the process data of the warning event according to the warning signal and store it in the cloud, and push the warning event and the process data to the law enforcement department to investigate and supervise the warning event.

[0010] Preferably, the method for preprocessing the cultural and tourism contract data includes:

[0011] Preprocess the contract data, remove missing, duplicate, and incorrect data, and standardize the price and schedule data;

[0012] Group and cluster the data. According to the destination information in the travel contracts, group the contract data by destination, perform cluster analysis on the contract data for each destination to identify similar groups of travel contracts, and use clustering algorithms to classify and divide the contracts according to contract price and itinerary;

[0013] Calculate the average price per person per day for each cluster group at each destination, and adjust the average price according to the off-peak and peak seasons of the travel destination. The calculation method of the average price is In the formula, Ap is the average price per person per day, Se is the seasonal adjustment ratio coefficient, Cp is the contract price, Nt is the number of tourists, Td is the number of days of the itinerary. Among them, the seasonal adjustment ratio coefficient is an impact factor that changes according to time and is obtained through time series analysis of the off-peak and peak season market conditions.

[0014] Preferably, the method for obtaining the seasonal adjustment ratio coefficient through time series analysis of the off-peak and peak season market conditions is as follows:

[0015] According to the contract signing date, divide the data into two groups: off-peak season and peak season, and select an appropriate time window to analyze the data;

[0016] Add a time index to the data for each time window, construct time series data, and perform smoothing processing on the time series data;

[0017] Use the time series decomposition method to decompose the time series data into trend components, seasonal components, and residual components, and based on the decomposition results, extract the seasonal components;

[0018] According to the characteristics of the time series, select the SARIMA model to analyze the data, fit the SARIMA model to the data, and optimize the model parameters by minimizing the residuals;

[0019] Use a part of the data for model training and another part of the data for model verification, and evaluate the performance of the model through residual analysis, AIC / BIC indicators, etc.;

[0020] Based on the trained model, predict the prices for future time windows, and obtain the seasonal adjustment ratio coefficient according to the results of the time series analysis.

[0021] Preferably, the time series decomposition method used is to perform decomposition operations on the components of the time series data using STL decomposition. The specific process is as follows:

[0022] After smoothing the time series, remove the seasonal components to obtain the initial trend components;

[0023] Subtract the initial trend component from the time series to obtain the initial seasonal component, and average the initial seasonal component within a period to obtain the initial seasonal component at each time point within each period;

[0024] Optimize the trend and seasonal components through multiple iterations. Each iteration includes extracting the seasonal component, extracting the trend component, and extracting the residual component;

[0025] Terminate the iteration according to the preset upper limit of the number of iterations or the convergence criterion, and output the final seasonal component, trend component, and residual component.

[0026] Preferably, the methods for extracting the seasonal component, extracting the trend component, and extracting the residual component in multiple iterations are as follows:

[0027] The method for extracting the seasonal component is to subtract the trend component from the time series to obtain the residual series, smooth each time point within each period in the residual series to obtain the smoothed seasonal component, and centralize the smoothed seasonal component so that its average value is zero;

[0028] The method for extracting the trend component is to subtract the seasonal component from the time series to obtain the deseasonalized series, and smooth the deseasonalized series to obtain the smoothed trend component;

[0029] The method for extracting the residual component is to subtract the trend component and the seasonal component from the original time series to obtain the residual component.

[0030] Preferably, the process of time series analysis according to the SARIMA model is as follows:

[0031] Use first-order differencing to eliminate the trend component, and use seasonal differencing with the period length to eliminate the seasonal component;

[0032] Plot the autocorrelation function and partial autocorrelation function graphs, identify the AR, MA, seasonal AR, and seasonal MA components in the function graphs, and select the SARIMA model parameters according to the autocorrelation function and partial autocorrelation function graphs;

[0033] Construct a SARIMA model according to the SARIMA model parameters, use an optimization algorithm to estimate the model parameters, fit the SARIMA model to the time series data, calculate the residuals of the model and evaluate the model performance;

[0034] Check the autocorrelation of the model residuals to ensure that the residuals are white noise, use statistical tests to verify the independence and normality of the residuals, and select the optimal model according to the information criterion;

[0035] Divide the data into a training set and a validation set, fit the model with the training set, evaluate the prediction performance of the model with the validation set, conduct rolling predictions, evaluate the prediction effect of the model at different time points, calculate the predicted values and confidence intervals, and provide the range of prediction uncertainty;

[0036] According to the diagnostic results and prediction performance, adjust the model parameters and conduct iterative optimization. After each parameter adjustment, refit the model and conduct validation until the best model is obtained.

[0037] Preferably, the method for constructing an abnormal contract detection model using machine learning algorithms to identify tourism contracts with abnormal prices is as follows:

[0038] Clean the tourism contract data to generate features, including geographical features, time features, traffic features, and price features, and standardize the features in numerical form and encode the categorical features;

[0039] Set the number of subsamples and the number of trees, input the generated feature data into the Isolation Forest model for training. The Isolation Forest constructs multiple isolation trees by randomly selecting features and split points to isolate data points, and calculates the anomaly score for each tourism contract by the Isolation Forest model;

[0040] Determine the preliminary anomaly score threshold according to the distribution of the anomaly scores, adjust the number of subsamples and the number of trees through cross-validation, and select the best parameter combination;

[0041] Set the anomaly score threshold, mark the contracts exceeding this threshold as abnormal contracts and send a warning signal.

[0042] Preferably, the logic for grading the warning signals according to the degree of unreasonable pricing is as follows:

[0043] In the automatic monitoring module, grade the warning signals according to the degree of unreasonable pricing. For each clustering group at each destination, limit the confidence interval of the reasonable price. Let the average price be Ap, then set the confidence floating value c%, and let the actual average price per person per day be Tr. When Tr is less than Ap(1 - 2c%), mark the tourism contract as a risk of an unreasonably low-price tour group and send a first-level risk signal;

[0044] When Tr is greater than or equal to Ap(1 - 2c%) and Tr is less than Ap(1 - c%), mark the tourism contract as a marginal risk of an unreasonably low-price tour group and send a second-level risk signal;

[0045] When Tr is greater than or equal to Ap(1 - c%), mark that the tourism contract has no risk of an unreasonably low-price tour group and send a healthy signal.

[0046] In the above technical solution, the technical effects and advantages provided by the present invention:

[0047] This application can centrally and collaboratively handle the chaos of additional charges disguised as unreasonably low prices in the industry based on cultural and tourism industry data. It can monitor the prices of tourism contracts in real time through an anomaly detection model, send risk signals through early warning methods, upload record process data as law enforcement evidence, use clustering analysis methods to classify consumer groups and consumption prices, analyze the characteristics of time series data in peak and off-peak seasons, extract seasonal impact factors affecting price fluctuations, and integrate information and share data among multiple departments to effectively solve the problems of data segmentation and information closure in the cultural and tourism industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0049] Figure 1 It is a system module diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0051] Embodiment 1

[0052] Please refer to Figure 1 As shown, the present invention is a big data-based intelligent cultural management collaborative management system, including a cultural data collection module, an analysis model construction module, an automatic monitoring module, and a collaborative response module;

[0053] The cultural data collection module is used to collect cultural and tourism contract data, and send the preprocessed cultural and tourism contract data to the analysis model construction module;

[0054] The analysis model construction module is used to construct a cultural and tourism mathematical model using data analysis technology and set a reasonable price threshold for the cultural and tourism mathematical model;

[0055] The automatic monitoring module is used to monitor newly signed cultural and tourism contracts in real time, construct an abnormal contract detection model using machine learning algorithms, identify tourism contracts with abnormal prices, and when it detects that the price of a cultural and tourism contract is lower than the reasonable price threshold set by the cultural and tourism mathematical model, it starts a warning signal and classifies the warning signal according to the degree of unreasonable pricing;

[0056] The collaborative response module is used to collect the process data of early warning events according to early warning signals and store them in the cloud, and push the early warning events and process data to the law enforcement department for investigation and supervision of the early warning events.

[0057] Cultural and tourism institutions may split the fees of low-cost tours into multiple parts or cover up the actual price through various hidden charges. For example, the low-cost tour fees may not include various necessary expenses such as catering, scenic spot tickets, transportation fees, etc., making the actual payment much higher than the advertised price.

[0058] The development and rise of the cultural and tourism industry have simultaneously introduced some chaotic phenomena caused by unfair competition into the industry. When cultural and tourism institutions sign contracts with tourists, they may use vague and unclear contract terms, making it difficult for tourists to detect or hold the unreasonable parts in the price accountable. There may be a large number of additional terms and fees in the contract, resulting in a huge gap between the actual payment and the advertised price.

[0059] Cultural and tourism institutions may sell the same tourism product at different prices through different sales channels. For example, they sell the same product through offline stores, online platforms, agents, etc., but the prices are different, resulting in the phenomenon of "different prices for the same tour group".

[0060] Some cultural and tourism institutions may take advantage of loopholes in current laws and regulations to design seemingly legal but actually unreasonable price strategies. For example, they classify certain fees as "self-funded items" or "voluntary activities" to avoid supervision.

[0061] By exaggerating publicity and hiding the actual situation, cultural and tourism institutions can take advantage of the information asymmetry of tourists to mislead tourists into choosing low-cost tours. Tourists often only discover various hidden fees and additional expenses during the actual travel process. However, since they have already paid and the itinerary has started, it is very difficult for tourists to safeguard their rights.

[0062] Some cultural and tourism institutions may package unreasonably low-cost tours as special offers or time-limited promotions to attract tourists to sign up. In fact, these so-called "offers" are just to attract customers, and the subsequent increase in various fees makes the overall cost much higher than expected.

[0063] The intelligent cultural management collaborative management system based on big data collects the cultural and tourism contract data of cultural and tourism institutions in a digitalized manner through the cultural data collection module. The collected data includes tourist information, tourist destinations, contract prices, itinerary days, etc. By collecting a large amount of tourism contract data, it can establish a comprehensive and accurate data foundation. As a prerequisite for subsequent analysis and modeling, the data can effectively reflect the actual situation and change trends of the market.

[0064] By analyzing the tourism contract data accumulated over the years, the laws and characteristics of the tourism market can be discovered and summarized, such as the average prices of different tourist destinations, price changes during peak and off-peak seasons, and the price composition of different types of tourist groups. Collecting and analyzing tourism contract data can build a mathematical model for identifying the reasonableness of prices. Through model calculation and analysis, the average price per person per day of a tourist destination can be obtained, and reasonable price thresholds can be set according to dimensions such as peak and off-peak seasons, tourist destination blocks, and source market. This is the basis for price early warning and an important basis for formulating early warning models and thresholds;

[0065] Through the analysis of contract data, abnormal contracts below the reasonable price threshold can be accurately identified. The system can automatically generate early warnings to remind law enforcement officers to conduct further manual verification, promptly discover and handle violations such as price ambiguity, and ensure the standardization and healthy development of the tourism market;

[0066] The collection and analysis of contract data provide strong data support for the cultural market law enforcement department. Through a data-driven supervision model, the efficiency and accuracy of law enforcement have been improved, making the supervision work more intelligent and refined;

[0067] Through scientific data analysis and model construction, industry price standards and norms can be gradually established. Some well-known online tourism platforms also use these standards as important references for the review of tourism product listings, promoting the standardized and healthy development of the entire tourism industry.

[0068] Based on cultural and tourism contract data, using data analysis techniques to construct a cultural and tourism mathematical model. The cultural and tourism mathematical model is used to analyze the average price level per person per day of a tourist destination. The process of constructing a model based on cultural and tourism contract data and relying on the model for monitoring and early warning is as follows;

[0069] Preprocess the contract data, remove missing, duplicate, and incorrect data, standardize price and schedule data, and extract features from the original contract data. The features include the number of tourists, number of travel days, tourist destination, season, and contract price;

[0070] Group and cluster the data. According to the destination information in the tourism contract, group the contract data by destination, conduct cluster analysis on the contract data of each destination, identify similar groups of tourism contracts, and use clustering algorithms to classify and divide the contracts according to contract price and itinerary;

[0071] Calculate the average price per person per day of each cluster group for each destination, and adjust the average price according to the peak and off-peak seasons of the tourist destination. The calculation method of the average price is Wherein, Ap is the average price per person per day, Se is the seasonal adjustment ratio coefficient, Cp is the contract price, Nt is the number of tourists, and Td is the number of days of the itinerary. Among them, the seasonal adjustment ratio coefficient is an impact factor that changes according to time and is obtained through time series analysis of the off-peak and peak season market conditions;

[0072] Use machine learning algorithms to construct an abnormal contract detection model to identify tourism contracts with abnormal prices. Use the contract data that has not participated in the construction of the abnormal contract detection model to verify the abnormal contract detection model, and adjust the model parameters according to the verification results to optimize the abnormal contract detection model;

[0073] Use the abnormal contract detection model to monitor newly signed tourism contracts. When an abnormal signal is detected, an early warning is issued by the intelligent cultural management collaborative management system.

[0074] Common standardization methods include min-max standardization, z-score standardization, fuzzy quantification method, etc.;

[0075] Common machine learning algorithms for anomaly detection include isolation forest, local outlier factor, etc.;

[0076] Common time series analysis methods include autoregressive moving average model, state space model, seasonal autoregressive integrated moving average model, long short-term memory network, recurrent neural network, etc.;

[0077] The seasonal adjustment ratio coefficient is obtained through time series analysis of the off-peak and peak season market conditions data. The process of time series analysis is as follows:

[0078] According to the contract signing date, divide the data into two groups: off-peak season and peak season, and select an appropriate time window to analyze the data;

[0079] Add a time index to the data of each time window, construct time series data, and smooth the time series data;

[0080] Use the time series decomposition method to decompose the time series data into trend components, seasonal components, and residual components. Based on the decomposition results, extract the seasonal components;

[0081] According to the characteristics of the time series, select the SARIMA model to analyze the data, fit the SARIMA model to the data, and optimize the model parameters by minimizing the residuals;

[0082] Use a part of the data for model training and another part of the data for model verification, and evaluate the performance of the model through residual analysis, AIC / BIC indicators, etc.;

[0083] Based on the trained model, predict the prices in the future time window, and obtain the seasonal adjustment ratio coefficient according to the results of time series analysis.

[0084] A part of the data for model training is the training set, and another part of the data for model validation is the validation set;

[0085] It should be noted that the grouping logic of the off-season and peak-season and the division of the time window are set by professionals in this field with reference to different tourist destinations;

[0086] Common smoothing methods include moving average, exponential smoothing, Holt linear trend method, local regression and other methods.

[0087] The time series decomposition method is to decompose the components of the time series data by using STL decomposition. The specific process is as follows:

[0088] After smoothing the time series, remove the seasonal component to obtain the initial trend component;

[0089] Subtract the initial trend component from the time series to get the initial seasonal component. Average the initial seasonal component within the period to get the initial seasonal component at each time point within each period;

[0090] Optimize the trend and seasonal components through multiple iterations. Each iteration includes extracting the seasonal component, extracting the trend component, and extracting the residual component;

[0091] Terminate the iteration according to the preset upper limit of the number of iterations or the convergence criterion, and output the final seasonal component, trend component, and residual component.

[0092] The method for extracting the seasonal component is to subtract the trend component from the time series to get the residual sequence. Smooth the time points within each period in the residual sequence to get the smoothed seasonal component, and centralize the smoothed seasonal component so that its average value is zero;

[0093] The method for extracting the trend component is to subtract the seasonal component from the time series to get the deseasonalized sequence, and smooth the deseasonalized sequence to get the smoothed trend component;

[0094] The method for extracting the residual component is to subtract the trend component and the seasonal component from the original time series to get the residual component.

[0095] The final seasonal component can reflect the periodic fluctuations of the time series, the trend component can reflect the long-term change trend of the time series, and the residual component can reflect the random fluctuations in the time series that do not belong to the trend or seasonality.

[0096] The advantage of STL decomposition lies in its ability to handle time series data with any period, being insensitive to outliers, having stable decomposition results, being able to clearly separate the trend, seasonal, and residual components of the time series, and having clear interpretability.

[0097] Select the SARIMA model to analyze the data according to the characteristics of the time series. The seasonal adjustment proportion coefficient with seasonal change characteristics is more suitable for analysis by the SARIMA model. The price of tourism contracts is often significantly affected by seasons. For example, the price difference between peak season and off-season is obvious. In the tourism industry, seasonal factors are particularly important for price fluctuations. The data shows regular fluctuation characteristics, such as higher or lower prices at specific times of each year. There is a long-term upward or downward trend in the data, such as the trend change of price due to changes in market demand or economic environment. There are random fluctuations in the data, and these fluctuations do not follow a clear pattern and need to be captured through the error term of the model, that is, the time series data of tourism contract prices has significant seasonal characteristics, trend characteristics, and random fluctuations.

[0098] The SARIMA model, also known as the seasonal ARIMA model, has significant advantages in dealing with time series data with seasonal and trend characteristics. The SARIMA model is specifically designed to capture the seasonal components in the time series. Seasonal autoregressive, seasonal differencing, and seasonal moving average terms are added to the model, enabling it to effectively handle periodic fluctuations. Through differencing, the SARIMA model can transform non-stationary time series data into stationary time series, making the model better able to fit and predict. The SARIMA model has multiple parameters that can be flexibly adjusted according to data characteristics, enabling the model to adapt to different trends and seasonal patterns.

[0099] Differencing includes non-seasonal differencing and seasonal differencing.

[0100] Although other time series models, including ARIMA, Exponential Smoothing, Prophet, etc., can also effectively handle time series data in some cases, SARIMA has more advantages in dealing with time series data with obvious seasonality and trends. For example, the ARIMA model is suitable for time series data without seasonality or insignificant seasonality, and Exponential Smoothing is suitable for time series data with stationarity or weak seasonal components, which may not be as effective as the SARIMA model in dealing with complex seasonality and long-term trends. The Prophet model is suitable for time series data with complex seasonality and holiday effects and is easy to use, suitable for dealing with data containing missing values and outliers. Although the Prophet model is suitable for dealing with complex seasonality, it is not as specific and refined as the SARIMA model in capturing pure seasonal fluctuations and trend changes. Therefore, the SARIMA model is chosen because it can handle both seasonal and trend components in the time series, providing more accurate fitting and prediction results.

[0101] The process of time series analysis according to the SARIMA model is as follows:

[0102] Use first-order differencing to eliminate the trend component and use periodic length differencing to eliminate the seasonal component;

[0103] Plot the autocorrelation function and partial autocorrelation function graphs, identify the AR, MA, seasonal AR, and seasonal MA components in the function graphs, and select the SARIMA model parameters according to the autocorrelation function and partial autocorrelation function graphs;

[0104] Construct the SARIMA model according to the SARIMA model parameters, use the optimization algorithm to estimate the model parameters, fit the SARIMA model to the time series data, calculate the residuals of the model and evaluate the model performance;

[0105] Check the autocorrelation of the model residuals to ensure that the residuals are white noise, use statistical tests to verify the independence and normality of the residuals, and select the optimal model according to the information criterion;

[0106] Divide the data into a training set and a validation set, fit the model with the training set, evaluate the model prediction performance with the validation set, perform rolling predictions, evaluate the prediction effect of the model at different time points, calculate the predicted values and confidence intervals, and provide the uncertainty range of the predictions;

[0107] According to the diagnostic results and prediction performance, adjust the model parameters, perform iterative optimization, and after each parameter adjustment, refit the model and perform validation until the best model is obtained.

[0108] Common optimization algorithms include maximum likelihood estimation, gradient descent, Bayesian optimization, genetic algorithms, etc. Common statistical test methods include Ljung-Box test, etc. Information criteria include AIC and BIC;

[0109] Use the Isolation Forest algorithm to construct an abnormal contract detection model to identify tourism contracts with abnormal prices. Machine learning algorithms are divided into two types: supervised learning and unsupervised learning. For problems of abnormal detection type, unsupervised learning is usually applied;

[0110] In the identification of price anomalies, there are few or no marked abnormal data. In tourism contracts, the number of cases that can clearly mark which contracts have abnormal prices is limited, making it difficult to provide sufficient training data for supervised learning. Anomalies are often rare events, and their characteristics are significantly different from normal data, but this difference is difficult to define in advance. Unsupervised learning algorithms discover these anomalies by analyzing the data itself without the need to define the specific characteristics of the anomalies in advance. The price fluctuations in the tourism market are affected by various factors such as season, destination, market demand, etc. Unsupervised learning algorithms can adaptively identify abnormal patterns in the data without relying on prior knowledge.

[0111] The Isolation Forest algorithm is an unsupervised learning algorithm specifically for anomaly detection. The time complexity of the Isolation Forest algorithm is linear. Compared with other algorithms that need to build dense models, the Isolation Forest is more efficient in processing large-scale data. The Isolation Forest constructs trees by randomly selecting features and split points, adapting to high-dimensional data without worrying about the problem of dimensionality disaster, which is particularly important for the various features that may be included in tourism contract data. The Isolation Forest does not assume data distribution and is very effective for anomaly detection. It identifies anomalies by isolating data points rather than based on a certain preset statistical distribution. The Isolation Forest algorithm isolates data points by constructing decision trees. The depth of the tree, that is, the path length, intuitively represents the degree of isolation of the data point. A shallower path means that the data point is easier to be isolated and is more likely to be an anomaly point. Due to the randomness of the algorithm, the Isolation Forest can improve the robustness of processing noisy data. Anomaly points are usually easier to be isolated than normal points and are not affected by individual noise points.

[0112] The construction process of the abnormal contract detection model is as follows:

[0113] Clean the tourism contract data to generate features, including geographical features, time features, traffic features, price features, and standardize the numerical features and encode the categorical features;

[0114] Set the number of subsamples and the number of trees, input the generated feature data into the Isolation Forest model for training. The Isolation Forest constructs multiple isolation trees by randomly selecting features and split points to isolate data points, and calculates the anomaly score for each tourism contract by the Isolation Forest model;

[0115] Determine the preliminary abnormal score threshold according to the distribution of abnormal scores, adjust the number of subsamples and the number of trees through cross-validation, and select the best parameter combination;

[0116] Set the abnormal score threshold, and mark the contracts exceeding this threshold as abnormal contracts and send early warning signals.

[0117] In the automatic monitoring module, the early warning signals are graded according to the degree of unreasonable pricing. A confidence interval for reasonable prices is defined for each cluster group at each destination. Taking the average price as Ap, a confidence floating value c% is set. Taking the actual average price per person per day as Tr, when Tr is less than Ap(1 - 2c%), mark the tourism contract as a risk of an unreasonably low-price group and send a first-level risk signal;

[0118] When Tr is greater than or equal to Ap(1 - 2c%) and Tr is less than Ap(1 - c%), mark the tourism contract as a marginal risk of an unreasonably low-price group and send a second-level risk signal;

[0119] When Tr is greater than or equal to Ap(1 - c%), mark the tourism contract as having no risk of an unreasonably low-price group and send a health signal.

[0120] Collect the process data of the early warning events according to the early warning signals and store them in the cloud. Push the early warning events and process data to the law enforcement department for investigation and supervision of the early warning events. The intelligent cultural management collaborative management system synchronizes the early warning information to departments such as the Internet information department, public security department, transportation department, and market supervision department for sharing. Each department can access the early warning information in real time through the system, obtain data, and subscribe to specific types of early warning information to obtain notification signals in a timely manner.

[0121] This application can centrally and collaboratively handle the chaos of extra charges disguised as unreasonably low prices in the industry based on cultural and tourism industry data. Real-time monitor the prices of tourism contracts through an anomaly detection model, send risk signals through early warning methods, upload record process data as law enforcement evidence, use clustering analysis methods to classify consumer groups and consumption prices, analyze the characteristics of time series data in peak and off-peak seasons, extract seasonal impact factors affecting price fluctuations, and integrate multiple departments for information connection and data sharing, effectively solving the problems of data fragmentation and information isolation in the cultural and tourism industry.

[0122] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data and performing software simulation to get a formula closest to the real situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0123] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired or wireless (such as infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that contains a collection of one or more available media. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, or a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0124] It should be understood that in various embodiments of the present application, the magnitudes of the sequence numbers of the above processes do not mean the order of execution, and the order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0125] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0126] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the system described above can refer to the corresponding process in the foregoing method embodiments, and will not be described herein again.

[0127] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of software products. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0128] As described above, the above are only specific implementation manners of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. The intelligent document management collaborative management system based on big data is characterized by: It includes cultural data collection module, analysis model building module, automatic monitoring module and collaborative response module; The cultural data collection module is used to collect cultural tourism contract data, and send the cultural tourism contract data to the analysis model construction module after pre-processing; The analysis model building module is used to build a cultural tourism mathematical model using data analysis technology and set a reasonable price threshold for the cultural tourism mathematical model; The automatic monitoring module is used to monitor newly signed cultural tourism contracts in real time, build an abnormal contract detection model using machine learning algorithms, and identify tourism contracts with abnormal prices. When it is detected that the price of a cultural tourism contract is lower than the reasonable price threshold of the set cultural tourism mathematical model, an early warning signal is activated and the early warning signal is graded according to the degree of unreasonable pricing. The collaborative response module is used to collect the process data of the warning event according to the warning signal and store it in the cloud, and push the warning event and process data to the law enforcement department to investigate and supervise the warning event; Methods for preprocessing cultural tourism contract data include: Pre-process contract data to remove missing, duplicate, and erroneous data, and standardize price and schedule data; Group and cluster the data. According to the destination information in the travel contract, group the contract data by destination, perform cluster analysis on the contract data of each destination, identify similar travel contract groups, and use clustering algorithms to classify and divide the contracts according to contract price and itinerary; Calculate the average price per person per day for each cluster group at each destination, and adjust the average price according to the off-season and peak-season conditions of the tourist destination. The average price is calculated as follows: In the formula, Ap is the average price per person per day, Se is the seasonal adjustment coefficient, Cp is the contract price, Nt is the number of tourists, Td is the number of days of the trip, among which the seasonal adjustment coefficient is the influencing factor that changes according to time, and is obtained by analyzing the time series of the off-season and peak-season market conditions; The method for obtaining the seasonal adjustment ratio coefficient based on the time series analysis of the off-season and peak-season market is: According to the contract signing date, the data is divided into two groups: off-season and peak season, and the appropriate time window is selected to analyze the data; Add a time index to the data in each time window, construct time series data, and perform smoothing on the time series data; Use the time series decomposition method to decompose the time series data into trend components, seasonal components and residual components, and extract the seasonal components based on the decomposition results; According to the characteristics of time series, the SARIMA model is selected to analyze the data, the SARIMA model is fitted to the data, and the model parameters are optimized by minimizing the residuals; Use part of the data for model training and the other part for model validation, and evaluate the performance of the model through residual analysis and AIC / BIC indicators; Based on the trained model, the price of the future time window is predicted, and the seasonal adjustment coefficient is obtained according to the results of time series analysis.

2. The intelligent document management collaborative management system based on big data according to claim 1 is characterized in that: The time series decomposition method uses STL decomposition to decompose the components of time series data. The specific process is: After smoothing the time series, the seasonal component is removed to obtain the initial trend component; Subtract the initial trend component from the time series to obtain the initial seasonal component, average the initial seasonal component within the cycle to obtain the initial seasonal component at each time point in each cycle; Optimize the trend and seasonal components through multiple iterations, each of which includes extracting the seasonal component, extracting the trend component, and extracting the residual component; The iteration is terminated according to the preset upper limit of iteration number or convergence standard, and the final seasonal component, trend component and residual component are output.

3. The intelligent document management collaborative management system based on big data according to claim 2 is characterized in that: The method of extracting seasonal components, trend components, and residual components in multiple iterations is: The method of extracting seasonal components is to subtract the trend component from the time series to obtain the residual series, smooth the time points in each cycle of the residual series to obtain the smoothed seasonal components, and center the smoothed seasonal components to make their average value zero; The method of extracting the trend component is to subtract the seasonal component from the time series to obtain a deseasonalized series, and then smooth the deseasonalized series to obtain a smoothed trend component. The method of extracting the residual component is to subtract the trend component and the seasonal component from the original time series to obtain the residual component.

4. The intelligent document management collaborative management system based on big data according to claim 3 is characterized in that: The process of time series analysis based on the SARIMA model is: Use first-order differences to eliminate trend components, and use period length differences to eliminate seasonal components; Draw autocorrelation function and partial autocorrelation function graphs, identify AR, MA, seasonal AR and seasonal MA components in the function graphs, and select SARIMA model parameters based on the autocorrelation function and partial autocorrelation function graphs; Construct a SARIMA model based on the SARIMA model parameters, estimate the model parameters using an optimization algorithm, fit the SARIMA model to time series data, calculate the model residuals and evaluate the model performance; Check the autocorrelation of model residuals, ensure that the residuals are white noise, use statistical tests to verify the independence and normality of residuals, and select the optimal model based on the information criterion; Divide the data into a training set and a validation set, fit the model with the training set, evaluate the model prediction performance with the validation set, perform rolling predictions, evaluate the prediction effect of the model at different time points, calculate the prediction value and confidence interval, and provide the uncertainty range of the prediction; According to the diagnosis results and prediction performance, the model parameters are adjusted and iterative optimization is performed. After each parameter adjustment, the model is refitted and verified until the optimal model is obtained.

5. The intelligent document management collaborative management system based on big data according to claim 1 is characterized in that: The method of using machine learning algorithms to build an abnormal contract detection model and identify travel contracts with abnormal prices is as follows: After cleaning the tourism contract data, features are generated, including geographical features, time features, traffic features, and price features. The features in numerical form are standardized and the categorical features are encoded. The number of subsamples and trees is set, and the generated feature data is input into the isolation forest model for training. The isolation forest constructs multiple isolated trees by randomly selecting features and partitioning points to isolate data points. The anomaly score of each travel contract is calculated by the isolation forest model. Determine the initial anomaly score threshold based on the distribution of anomaly scores, adjust the number of subsamples and the number of trees through cross-validation, and select the best parameter combination; Set an abnormal score threshold, mark contracts that exceed the threshold as abnormal contracts, and send early warning signals.

6. The intelligent document management collaborative management system based on big data according to claim 5 is characterized in that: The logic for grading early warning signals according to the degree of unreasonable pricing is: In the automatic monitoring module, the early warning signals are graded according to the degree of unreasonable pricing. Each cluster group of each destination defines the confidence interval of reasonable prices. The average price is Ap, and the confidence floating value c% is set. The actual average price per person per day is Tr. When Tr is less than Ap (1-2c%), the travel contract is marked as an unreasonable low-price group risk and a first-level risk signal is sent. When Tr is greater than or equal to Ap(1-2c%) and Tr is less than Ap(1-c%), the travel contract is marked as an unreasonable low-price group marginal risk and a secondary risk signal is sent; When Tr is greater than or equal to Ap(1-c%), the travel contract is marked as having no risk of unreasonable low-price tours, sending a healthy signal.

Images