Railway freight market potential evaluation method and device, and electronic equipment

Abstract

The invention provides a railway freight market potential evaluation method and apparatus, and an electronic device. The method comprises the steps of obtaining historical freight market data and a customer portrait; inputting the historical freight market data and the customer portrait into a potential evaluation model constructed based on a neural network to obtain a subject evaluation result and a corrected evaluation result; and obtaining a market potential evaluation result based on the subject evaluation result and the corrected evaluation result. According to the method and the device provided by the invention, the historical freight market data and the customer portrait are respectively input into the potential evaluation model constructed based on the neural network, the subject evaluation result and the corrected evaluation result output by the potential evaluation model are obtained, and the customer portrait is evaluated based on the subject evaluation result and the corrected evaluation result. The accuracy is higher, the reference value is larger, and the railway freight enterprises can be effectively helped to optimize transport capacity resource allocation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of railway freight market evaluation, and particularly relates to a railway freight market potential evaluation method, device and electronic equipment. BACKGROUND

[0002] The commonly used market potential evaluation method and system generally collects basic data of market potential evaluation of a to-be-evaluated region, takes a comprehensive energy consumption benchmark value deviation rate, a comprehensive energy consumption average value deviation rate and a total value average growth rate as characteristic values, calculates weight coefficients of the characteristic values based on a principal component analysis method, and finally calculates a specific business market potential value according to the characteristic values and weights, and determines the size of the business potential according to the potential value.

[0003] However, the current principal component analysis method highly depends on the selection quality of characteristic variables and the rationality of weight distribution, and the accuracy and practicability of the potential evaluation result need to be further improved. SUMMARY

[0004] The present application provides a railway freight market potential evaluation method, device and electronic equipment to solve the defects of the principal component analysis based on the principal component analysis in the prior art.

[0005] The present application provides a railway freight market potential evaluation method, which comprises the following steps: acquiring historical freight market data and customer portraits; inputting the historical freight market data and the customer portraits into a potential evaluation model respectively to obtain a subject evaluation result and a correction evaluation result output by the potential evaluation model; obtaining a market potential evaluation result based on the subject evaluation result and the correction evaluation result; The potential evaluation model is constructed based on a neural network.

[0006] According to the railway freight market potential evaluation method provided by the present application, the historical freight market data and the customer portraits are input into the potential evaluation model respectively to obtain a subject evaluation result and a correction evaluation result output by the potential evaluation model, which comprises the following steps: inputting the historical freight market data and the customer portraits into the potential evaluation model respectively; applying the historical freight market data to turnover volume prediction and line profit prediction based on the potential evaluation model to obtain the subject evaluation result; applying the customer portraits to output the correction evaluation result based on the potential evaluation model.

[0007] According to the method for evaluating the potential of the railway freight market provided by the present invention, the step of applying the customer profile based on the potential evaluation model and outputting the revised evaluation result includes: Extract customer profile features from the customer profile; The revised evaluation result is obtained by applying the customer profile features based on the potential evaluation model.

[0008] According to a method for evaluating the potential of the railway freight market provided by the present invention, the steps for obtaining historical freight market data include: Collect initial historical freight market data; The initial historical freight market data is preprocessed to obtain clean market data; Perform a data verification operation on the cleaning market data to obtain the historical freight market data; The data verification operation includes at least one of data integrity verification, logical consistency verification, and outlier detection.

[0009] According to the method for evaluating the potential of the railway freight market provided by the present invention, the historical freight market data includes at least one of macroeconomic data, industry traffic data, publicly available market data, and railway-enterprise interaction data; The customer profile includes at least one of the following: customer industry attributes, frequency of cargo transportation, volume of transportation, payment ability, and historical stability of cooperation.

[0010] According to a method for evaluating the potential of a railway freight market provided by the present invention, the step of obtaining a market potential evaluation result based on the main evaluation result and the revised evaluation result includes: Based on a supervised learning algorithm, abnormal results in the market potential evaluation results are analyzed. If the abnormal result exceeds the warning threshold, the parameters of the potential evaluation model are adjusted.

[0011] The present invention also provides an evaluation device for the potential of the railway freight market, comprising: The acquisition unit retrieves historical freight market data and customer profiles. The evaluation unit inputs the historical freight market data and the customer profile into the potential evaluation model to obtain the main evaluation result and the revised evaluation result output by the potential evaluation model. The fusion unit obtains the market potential evaluation result based on the main evaluation result and the revised evaluation result; The potential evaluation model is constructed based on a neural network.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for evaluating the potential of the railway freight market as described above.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for evaluating the potential of the railway freight market as described above.

[0014] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the method for evaluating the potential of the railway freight market as described above.

[0015] The method, apparatus, and electronic equipment for evaluating the potential of the railway freight market provided by this invention input historical freight market data and customer profiles into a potential evaluation model constructed based on a neural network, respectively, to obtain the main evaluation result and the corrected evaluation result output by the potential evaluation model. Based on the main evaluation result and the corrected evaluation result, a more accurate and valuable reference can be obtained, which can effectively help railway freight companies optimize the allocation of transport capacity resources. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is one of the flowcharts illustrating the method for evaluating the potential of the railway freight market provided by this invention; Figure 2 This is the second flowchart illustrating the method for evaluating the potential of the railway freight market provided by this invention. Figure 3 This is a schematic diagram of the structure of the railway freight market potential evaluation device provided by the present invention; Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0019] To address the aforementioned problems, this invention provides a method for evaluating the potential of the railway freight market, thereby achieving accurate and highly practical analysis of the railway freight market potential. Figure 1 This is one of the flowcharts illustrating the method for evaluating the potential of the railway freight market provided by this invention, such as... Figure 1 As shown, the method includes: Step 110: Obtain historical freight market data and customer profiles.

[0020] Here, historical freight market data refers to a set of structured or unstructured data reflecting the objective occurrence of railway freight operations within a specific historical time period. Historical freight market data is primarily used to reveal the market's development trends, cyclical fluctuations, and fundamental patterns at the macro level.

[0021] Furthermore, the customer profile here refers to the multi-dimensional characteristics of freight customers, including demographic information, transaction behavior, preferences, and potential needs. Customer profiling is primarily used to understand customer behavior patterns and potential needs at a micro level, filling gaps in identifying potential market opportunities that cannot be discovered using only historical data.

[0022] Specifically, comprehensive historical freight market data can be extracted from internal railway enterprise systems such as transportation management, finance, and enterprise resource planning, or obtained from external public channels such as industry associations and third-party data platforms. This data can provide factual basis for subsequent evaluation of freight market potential. Historical freight market data here may include, but is not limited to, waybill data, such as cargo category, weight, volume, volume, freight rate, distance, and transit time; spatiotemporal data, such as the time of the transport operation, origin, and destination; and macroeconomic data, such as industry GDP, import and export data, and commodity price indices related to specific goods.

[0023] Additionally, customer profiles can be obtained through customer relationship management systems, customer interview records, market research questionnaires, and information mining from publicly available online sources, or from pre-built customer profile ratings. Customer profiles may include, but are not limited to, basic customer information such as the customer's industry, company size, registered capital, and geographical location; historical transaction behavior such as shipping frequency, duration of cooperation, historical freight category preferences, average order value, and the date of the most recent transaction; service preference characteristics such as sensitivity to transportation timeliness, price sensitivity, and demand for value-added services such as warehousing; and potential demand characteristics such as potential new cargo sources or new routes derived from analyzing the customer's company news, bidding information, and industry development trends.

[0024] It should be noted that by acquiring historical freight market data and customer profiles, the potential evaluation model can learn general market patterns and gain insights into potential growth points driven by specific customer needs, thus laying a solid data foundation for the comprehensiveness and accuracy of the final evaluation results.

[0025] Step 120: Input the historical freight market data and the customer profile into the potential evaluation model to obtain the main evaluation result and the corrected evaluation result output by the potential evaluation model; the potential evaluation model is constructed based on a neural network.

[0026] Here, the main evaluation result refers to the benchmark and trend evaluation of future market potential output by the potential evaluation model, which is primarily based on the macroeconomic trends and cyclical patterns of historical freight market data. This result can be understood as the market's potential prediction based on its historical inertia, under ideal conditions where individual customer needs are not considered.

[0027] The revised evaluation result here refers to a supplementary evaluation that dynamically adjusts the main evaluation result by combining the potential evaluation model with the individualized and differentiated characteristics in the customer profile. This result aims to quantify the market potential brought about by specific customer behaviors, preferences, or potential needs. For example, when the customer profile shows that a certain customer group has increased requirements for timeliness and may be willing to pay higher fees for freight, the revised evaluation result might be a positive adjustment coefficient or incremental value. This could be a weighting coefficient, a specific increase or decrease, or a revised probability.

[0028] Furthermore, the potential evaluation model in this embodiment is based on a neural network. It is understood that the specific structure of the neural network is not uniquely limited and can be various network structures well-known to those skilled in the art. For example, a multilayer perceptron can be used as the basic model; considering the temporal characteristics of historical data, a recurrent neural network or its variants, such as a long short-term memory network or a gated recurrent unit, can also be used. In a preferred embodiment of the invention, a multi-task learning neural network model can be constructed, which includes a shared feature extraction layer and two independent output layers: one output layer outputting the main evaluation result, and the other output layer outputting the corrected evaluation result.

[0029] Specifically, the historical freight market data and customer profiles acquired and preprocessed in the previous step can be input into a pre-built potential evaluation model for calculation. This model, built on neural network technology, is capable of learning complex nonlinear relationships within the data. From this, the main evaluation result and the revised evaluation result output by the potential evaluation model can be obtained.

[0030] It should be noted that by employing a neural network-based potential evaluation model, deep and non-linear complex relationships between data can be effectively uncovered, which is difficult to achieve with traditional linear models or simple statistical methods. More importantly, by decoupling the model's output into the main evaluation result and the revised evaluation result, the evaluation process becomes more transparent and interpretable. It clearly distinguishes between the market's basic potential and incremental potential, enabling business personnel not only to know the magnitude of the potential but also to understand its sources, thus providing richer insights for subsequent decision-making.

[0031] Step 130: Based on the subject evaluation results and the revised evaluation results, obtain the market potential evaluation results.

[0032] Here, the market potential evaluation result refers to a comprehensive and accurate quantitative or qualitative assessment of the future freight potential of a specific market, which integrates the macro-trends of the main evaluation results and the micro-dynamics of the revised evaluation results. For example, it may form a qualitative assessment for a specific route, product category, or customer. Understandably, the market potential evaluation result can serve as a direct basis for guiding railway freight departments in marketing, resource allocation, and product design. This includes optimizing capacity resource allocation, identifying high-value potential customers, and designing personalized freight products, thereby improving market responsiveness and overall competitiveness.

[0033] Specifically, the main evaluation results and the revised evaluation results obtained in the previous step are weighted and integrated to generate the final, comprehensive market potential evaluation results.

[0034] It should be noted that the fusion method can be varied. In one specific embodiment, fusion can be performed using a weighted summation method. For example, the market potential evaluation result = α Main evaluation results + β The revised evaluation results use α and β as weighting coefficients, which can be pre-set based on experience, for example, α=0.7 and β=0.3; or they can be adaptively learned by the model based on the importance of the data. In another embodiment, the fusion rules can be more complex, for example, implemented through a simple rule engine. For instance, when the corrected evaluation result exceeds a certain threshold, the main evaluation result is increased by a specific percentage; otherwise, it is decreased or remains unchanged.

[0035] It should be noted that by organically integrating the main evaluation results and the revised evaluation results, an effective combination of macro trends and micro dynamics is achieved. This ensures that the final market potential evaluation results not only reflect the overall development direction of the market but also sensitively capture structural opportunities and risks brought about by changes in customer demand, thereby greatly improving the accuracy, timeliness, and operability of the evaluation results.

[0036] The method provided in this invention inputs historical freight market data and customer profiles into a potential evaluation model built on a neural network, and obtains the main evaluation result and the corrected evaluation result output by the potential evaluation model. Based on the main evaluation result and the corrected evaluation result, a more accurate and valuable result is obtained, which can effectively help railway freight companies optimize the allocation of transport capacity resources.

[0037] Based on any of the above embodiments, step 120 includes: The historical freight market data and the customer profile are respectively input into the potential evaluation model; Based on the potential evaluation model, the historical freight market data is used to predict turnover and route profit, and the main evaluation results are obtained. Based on the potential evaluation model, the customer profile is applied to output the revised evaluation result.

[0038] Specifically, firstly, historical freight market data and customer profiles can be input into the potential evaluation model through two separate input branches.

[0039] Then, within the internal structure of the potential evaluation model, features extracted from the input branch of historical freight market data are fed to one or more output layers to generate the main evaluation results. The output layers are trained to perform two specific regression tasks: predicting future turnover and predicting future route profits. Therefore, the potential evaluation model outputs structured data consisting of two specific predicted values ​​with clear business implications. For example, for the potential evaluation of the "Beijing-Guangzhou" route in the first quarter of 2026, the main evaluation result could be "{Forecasted turnover: 5 billion ton-kilometers, Forecasted route profit: 80 million yuan}".

[0040] It should be noted that turnover forecast refers to the prediction of freight turnover for a specific railway freight line or region within a future time period. It is a core indicator for measuring the workload of transportation companies and directly reflects market activity and transport scale. Additionally, line profit forecast refers to the prediction of the expected profitability of a specific railway freight line within a future time period. This forecast is typically calculated based on predicted revenue and predicted costs. Specifically, line profit can be obtained from "(predicted freight rate × predicted freight volume) - predicted total cost," where predicted total cost can be further broken down into transportation costs, station operation costs, labor costs, depreciation costs, etc.

[0041] It should be noted that by specifying the main evaluation results into two core operating indicators—turnover volume and route profit—the evaluation results are directly linked to the company's operational decisions, greatly enhancing the operability and business guidance value of the evaluation results. As a result, managers can clearly identify which routes are high-turnover, high-profit freight routes and which are high-turnover, low-profit, low-margin routes, thereby enabling differentiated resource allocation and marketing strategy development.

[0042] Similarly, within the internal structure of the potential assessment model, the features extracted from the customer profile input branch are fed to a separate output layer for generating revised assessment results. This output layer can be trained to output various forms of revised assessment results. These results can be revision coefficients, revision amounts, or revision categories / probabilities. For example, when the revision result is a revision coefficient, the model can output one or more multiplicative coefficients. For instance, outputting a profit revision coefficient of 1.1 indicates that, based on the customer profile, the predicted route profit has a 10% upward revision potential. Another example is a revision amount. When the revision result is a revision amount, the model can directly output an additive term. For example, outputting a turnover volume revision value of "+5 million ton-kilometers" indicates that the potential demand from this type of customer is expected to bring additional transportation growth. Yet another example is a positive category / probability; the model can output a classification result, such as high growth potential, service upgrade opportunity, or churn risk, along with corresponding probabilities.

[0043] It should be noted that by applying customer profiles to generate revised evaluation results, the method provided in this embodiment of the invention can quantify soft factors such as customers' personalized needs, industry development trends, and customer loyalty into direct impacts on hard operating indicators such as turnover and profit. Therefore, the final potential evaluation is no longer static and lagging, but dynamic, forward-looking, and personalized, helping companies capture structural market growth driven by customer needs and achieve truly precise marketing and in-depth customer cultivation.

[0044] The method provided in this invention, through a neural network structure with dual-branch input and multi-task output, organically combines turnover and profit forecasts based on historical freight market data with micro- and dynamic adjustments based on customer profiles. This not only improves the model's predictive accuracy but, more importantly, significantly enhances the interpretability and operability of the evaluation results. Decision-makers receive not a single, general potential score, but a composite evaluation report that includes basic market forecasts and dynamic adjustments based on customers. This allows for a deeper understanding of the composition of market potential and the development of more precise and effective market development and customer maintenance strategies.

[0045] Based on any of the above embodiments, and based on the potential evaluation model, the customer profile is applied to output the revised evaluation result, including: Extract customer profile features from the customer profile; The revised evaluation result is obtained by applying the customer profile features based on the potential evaluation model.

[0046] Here, customer profile features refer to standardized quantitative indicators or classification labels that are extracted, processed, or calculated from the original customer profile data and can characterize the characteristics of a customer in a certain dimension.

[0047] Specifically, firstly, the original customer profile, which may contain a large amount of unstructured or descriptive information, is processed and transformed into structured and quantitative variables that can be directly understood and calculated by the potential evaluation model, that is, the customer profile features are extracted.

[0048] In detail, the process of extracting customer profile features can select an encoding method that matches the data type of the customer profile, including but not limited to the following methods: Method 1: Labeling and Encoding; For categorical information, numerical encoding is used. For example, the industry to which a customer belongs is one-hot encoded, converting it into a multi-dimensional 0 / 1 vector; the customer size is encoded with ordinal numbers, such as 3, 2, 1 respectively.

[0049] Method two: Numericalization and normalization; For numerical information, standardization is performed to eliminate the influence of dimensions. For example, for numerical values ​​such as a company's registered capital and annual turnover, Z-score standardization or maximum-minimum normalization can be used to scale them to a fixed range, such as 0 to 1 or -1 to 1.

[0050] Method three involves constructing composite features. These features can be calculated based on multiple fundamental pieces of information to provide deeper business insights. For example, based on a customer's historical transaction behavior, the time of the most recent transaction, transaction frequency, and transaction amount can be calculated. The combination of these three features effectively measures customer value and loyalty. Another example is analyzing customer complaint rates regarding delays and whether they frequently choose expedited services to quantify price and time sensitivity, extracting service sensitivity features. Furthermore, natural language processing technology can be used to analyze customer-related industry news, publicly available bidding information, annual reports, and other textual data to extract keywords such as "expansion," "new base," and "supply chain upgrade," and based on this, calculate the customer's potential freight demand intensity score, extracting potential demand intensity features.

[0051] Next, customer profile features can be input into the customer profile input branch of the potential evaluation model. After receiving this feature vector, the branch uses a series of internal nonlinear transformations—namely, neurons and activation functions—to perform complex combinations and weights on these features, learning the nonlinear impact of different features on market potential. Finally, the output layer of this branch will output one or a set of numerical values, which represent the corrected evaluation result.

[0052] Based on any of the above embodiments, the steps for obtaining historical freight market data include: Collect initial historical freight market data; The initial historical freight market data is preprocessed to obtain clean market data; Perform a data verification operation on the cleaning market data to obtain the historical freight market data; The data verification operation includes at least one of data integrity verification, logical consistency verification, and outlier detection.

[0053] Specifically, the first step is to collect initial historical freight market data. This collection can be performed using automated scripts or data integration tools, such as ETL (Extract-Transform-Load) tools. Here, initial historical freight market data refers to the raw data record set directly extracted from various business systems or external channels, without having undergone cleaning, verification, or integration. This data represents the most primitive form of information; it can be comprehensive but of varying quality, potentially containing significant noise, errors, missing data, and inconsistencies.

[0054] Next, the initial historical freight market data can be processed, which may include one or more of the following operations. For example, handling missing values, i.e., for empty fields in the data records, using strategies such as deleting records, filling with the mean / median / mode, or for time series data, filling with preceding and following values, depending on the importance of the field and the proportion of missing values.

[0055] For example, standardizing data formats involves unifying data from different sources into a standard format. This includes unifying dates such as "October 10, 2024" and "10 / 10 / 2024" into the standard "YYYY-MM-DD" format; and converting weight units like "ton," "kg," and "kilogram" to "ton." Another example is removing duplicate records, which involves identifying and deleting identical waybill or customer records to avoid data redundancy caused by errors during data collection or merging.

[0056] It is understandable that clean market data refers to a dataset obtained after basic cleaning operations such as missing value handling, format standardization, and duplicate removal, based on initial historical freight market data. This dataset is regular and consistent in data structure and format, but may still contain logically unreasonable or out-of-range values.

[0057] Furthermore, data validation operations are performed on the clean market data to obtain historical freight market data. Here, data validation operations include at least one of data integrity validation, logical consistency validation, and outlier detection. Specifically, data integrity validation may involve checking whether core fields are complete. For example, a rule can be set that a valid waybill record must simultaneously contain four fields: "shipping date," "originating station," "destination station," and "cargo weight." Any record lacking any one of these fields will be considered incomplete data and marked or removed.

[0058] Logical consistency checks can be used to verify whether there are logical contradictions between data. For example, validation rules can be written to verify this, such as: For time-series logic: the "goods receipt time" of a waybill record must be later than the "shipment time". For business logic: goods marked as "cold chain transportation" should not be classified as "coal" or "steel". For spatial logic: for waybills from "Beijing" to "Tianjin", the "transportation mileage" should be within a reasonable range, such as 100-150 kilometers, and not 10 kilometers or 1000 kilometers.

[0059] Outlier detection refers to identifying and handling data points that deviate extremely from the general population. Outlier detection can be implemented using statistical methods, such as the 3-sigma rule or the IQR rule of box plots, to identify records where unit price, freight volume, or transportation time values ​​are significantly higher or lower than the normal range. These records may be due to data entry errors. Alternatively, outlier detection can also be implemented using model-based methods. For more complex multidimensional data, algorithms such as Isolation Forest can be used to detect combinations that do not conform to normal data patterns, such as a non-express freight order with an extremely short distance but an extremely high freight rate.

[0060] It should be noted that problematic data identified during verification can be handled by methods such as removal, marking, or contacting the business department for manual verification and correction.

[0061] The method provided in this invention effectively removes misleading erroneous data that cannot be detected by format cleaning alone, preventing the model from learning incorrect patterns from erroneous data. This results in historical freight market data with high accuracy and reliability, providing the strongest guarantee for training a robust and accurate potential evaluation model and fundamentally improving the reliability of the final results of the entire evaluation method.

[0062] Based on any of the above embodiments, historical freight market data includes at least one of macroeconomic data, industry traffic data, publicly available market data, and road-enterprise interaction data; The customer profile includes at least one of the following: customer industry attributes, frequency of cargo transportation, volume of transportation, payment ability, and historical stability of cooperation.

[0063] Based on any of the above embodiments, step 130 is followed by: Based on a supervised learning algorithm, abnormal results in the market potential evaluation results are analyzed. If the abnormal result exceeds the warning threshold, the parameters of the potential evaluation model are adjusted.

[0064] Specifically, after the model evaluates the market potential for a future time period, it's necessary to collect actual business data from the company's business system after that period ends. This data includes actual turnover and line profits generated, to gather real business results. Then, the model's market potential evaluation results are matched against the collected actual business results one by one. By calculating the difference between the two and comparing it to a preset error threshold, each evaluation record is labeled as "normal" or "abnormal." These labeled records collectively constitute a labeled dataset used to train the supervised learning algorithm.

[0065] Then, a supervised learning algorithm can be used to train the anomaly analysis model. The task of this algorithm is not to predict market potential, but rather to learn under what conditions the potential assessment model's predictions will become inaccurate. Here, supervised learning algorithms include, but are not limited to, decision trees or random forests, support vector machines, etc.

[0066] When abnormal results exceed the warning threshold, the parameters of the potential evaluation model are adjusted. Specific methods for parameter adjustment can include incremental learning or fine-tuning. For example, newly collected real data can be used as a training batch, and several more rounds of training can be conducted based on the existing potential evaluation model. This method typically uses a low learning rate, aiming to retain the model's existing knowledge while selectively learning new patterns and correcting errors.

[0067] It can also be model retraining. That is, if the model performance deteriorates significantly, or if the amount of new data accumulated is large enough, the new data can be merged with all historical data to perform a complete retraining of the potential evaluation model.

[0068] Alternatively, the weights of the loss function can be adjusted. That is, during retraining or fine-tuning, samples that are identified as anomalous results can be given higher weights, forcing the model to pay more attention to these samples that were previously predicted incorrectly during training, thereby enabling more targeted learning.

[0069] The method provided in this invention significantly enhances the robustness of the entire railway freight market potential evaluation method by introducing an analysis-early warning-adjustment closed-loop feedback loop. In other words, it transforms the model from a one-off predictive tool into a continuously learning and iterative intelligent system. This not only ensures the continuous accuracy of the evaluation results but, more importantly, establishes an automated mechanism for monitoring and maintaining the model's own health, reducing long-term manual maintenance costs and ensuring that the method remains advanced and valuable for decision-making in a dynamically changing market.

[0070] Based on any of the above embodiments Figure 2 This is the second flowchart illustrating the method for evaluating the potential of the railway freight market provided by this invention. Figure 2 As shown, the method includes: Step 1: Multi-channel data collection. Specifically, macroeconomic data, industry traffic data, publicly available market data, and data exchange between road companies and transportation companies are used as inputs. Through API integration, ETL toolchains, and streaming data processing technologies, relevant data is collected from multiple sources to obtain the raw dataset.

[0071] Step Two: Data Cleaning and Standardization. In detail, the original dataset is used as input, and data quality monitoring tools and techniques are used to detect outliers, missing values, and duplicate values. Data validation mechanisms are applied to ensure data accuracy and consistency. Data cleaning and standardization steps are performed to remove invalid data and convert the data to a uniform format. The final output is a structured dataset that conforms to the data lake's inbound standards. It is understandable that by detecting and handling outliers, missing values, and duplicate values, data quality is significantly improved.

[0072] Step 3: Data Lake Storage. Specifically, the cleaned and standardized data is used as input and stored in a data lake for subsequent feature extraction and model computation.

[0073] Step 4: Feature Extraction and Model Calculation. In detail, for the data in the data lake, relevant features are extracted from the data, such as customer industry, goods category, and transportation mode. For the structured data in the data lake, the Pandas library is used to read the Parquet format data. Categorical feature fields such as customer industry, goods category, and transportation mode are accurately extracted. The `map` function is then used to map transportation modes to numerical codes (e.g., railway → 1, highway → 2). Missing values ​​are uniformly filled using `fillna('unknown')`. Finally, a lightweight feature dataset is output using the `to_feather` method, ensuring that the feature extraction process conforms to data lake storage specifications and has high reproducibility.

[0074] Step 5: Customer Profile Rating. In detail, using the feature data as input, customer profiles are constructed and rated based on multi-dimensional customer information, outputting customer rating results. Using multi-dimensional information from the feature data, such as customer industry attributes, freight frequency, freight volume, payment ability, and historical cooperation stability, the customer group is segmented using clustering analysis algorithms to construct differentiated profiles. The analytic hierarchy process (AHP) is then used to assign weights to each dimension and calculate a comprehensive score. Finally, based on preset rating criteria, customer rating results are output, achieving a quantitative mapping from feature data to customer value levels.

[0075] Step Six: Evaluation Model Application. In detail, customer data is input into the evaluation model to assess market potential and output the market potential assessment results. The evaluation model is constructed using a neural network.

[0076] First, feature engineering is performed on the input customer data—structured data such as customer rating results, historical freight volume, transportation frequency, and regional economic indicators are processed into categorical variables through one-hot encoding, and continuous variables are standardized using Z-Score. Then, based on the evaluation model built in step four, grid search is used to tune the parameters to determine the optimal hyperparameters. In the model application stage, K-fold cross-validation is used to ensure generalization ability, and SHAP values ​​are used to interpret feature contribution. Finally, the market potential assessment results, which include potential freight volume growth space, market share, and demand response elasticity, are output, and a visual heatmap and trend prediction curve are generated as decision support appendices.

[0077] Taking the output process of the predicted railway freight volume branch as an example: First, determine whether the historical freight duration of the sample exceeds one year. If it does not exceed one year, calculate: the total historical freight volume / number of months. Use this result as the final result. If it exceeds one year, proceed to the following steps: use a BP neural network for modeling. That is, first, take the Log10 value of the historical freight volume data for normalization. Then, transform the data structure using a sliding window. Next, model a BP feedforward neural network for each sliding window. The hidden layer is set to two layers; the first hidden layer is set to 1 + the sliding window order / 2 neurons; the second hidden layer is set to 1 + the number of neurons in the first hidden layer / 2 neurons; the activation function of the first hidden layer is a linear function; the activation function of the second hidden layer is a sigmoid function; the iterative algorithm is the Levenberg-Marquardt method. Then, calculate the fitting residual under each sliding window. Finally, calculate the predicted value and perform back-normalization.

[0078] Furthermore, GBRT modeling is employed. That is, data structure transformation is performed using a sliding window. GBRT modeling is applied for each sliding window type. No feature engineering is performed; the data is directly input into the model. There are 5000 trees, each with a maximum of 6 leaf nodes; the learning iteration step size is 0.005. The fitting residuals under each sliding window type are calculated. Then, the predicted values ​​are calculated. The fitting residuals of all the above sub-models are compared, and the model with the smallest residual is selected as the final model. Correspondingly, the prediction result is the final prediction result. Finally, the prediction result is rounded; if the predicted freight volume is negative, it is directly set to zero.

[0079] Step 7: Supervised Learning Control and Early Warning. In detail, based on the market potential assessment results, a supervised learning algorithm is applied to analyze abnormal patterns in the assessment results. Early warning thresholds can be set and managed; when the results exceed a preset range, an early warning is triggered to obtain the final early warning information and adjustment suggestions.

[0080] Step 8: Visualization. Specifically, data visualization tools can be used to graphically display the assessment results and early warning information. The output can be a visual report or a dashboard. Step 9: Feedback and Optimization. In detail, based on the feedback information, the data collection, cleaning, and model calculation steps are optimized and adjusted to obtain an optimized process and model.

[0081] The method provided in this invention extracts characteristic indicators that meet the potential market analysis of railway freight based on multi-dimensional data collection: revenue performance, revenue contribution, profit, freight rate, competitor benchmarking, service quality, customer loyalty, market share, vehicle / equipment utilization rate, route conditions, capacity matching, manpower efficiency, market risk, industry fluctuations, policy subsidies, and other indicator characteristics used for potential analysis of the railway freight market.

[0082] Furthermore, a railway freight potential market analysis model was established to facilitate comprehensive analysis of railway freight potential. This model enables research on comprehensive analysis and evaluation of railway freight market potential. Based on the needs of freight market monitoring and marketing, market monitoring data processing technology is used, combined with customer profiles and ratings, to establish a multi-level comprehensive analysis and evaluation model of railway freight market potential, including customer industry, cargo category, and transportation mode. This model analyzes the possibility and risks of mutual conversion between other transportation modes and railway transportation modes for enterprises, and relies on machine learning, large-scale modeling, and other technologies to study the improvement, iteration, and automatic evolution mechanism of model algorithms.

[0083] Based on any of the above embodiments Figure 3 This is a schematic diagram of the structure of the railway freight market potential evaluation device provided by the present invention, as shown below. Figure 3 As shown, the device includes: Unit 310 retrieves historical freight market data and customer profiles; Evaluation unit 320 inputs the historical freight market data and the customer profile into the potential evaluation model to obtain the main evaluation result and the revised evaluation result output by the potential evaluation model. The fusion unit 330 obtains the market potential evaluation result based on the main evaluation result and the revised evaluation result; The potential evaluation model is constructed based on a neural network.

[0084] The device provided in this invention inputs historical freight market data and customer profiles into a potential evaluation model built on a neural network, and obtains the main evaluation result and the corrected evaluation result output by the potential evaluation model. Based on the main evaluation result and the corrected evaluation result, the device can obtain more accurate and more valuable results, and can effectively help railway freight companies optimize the allocation of transport capacity resources.

[0085] Based on any of the above embodiments, the evaluation unit is specifically used for: The historical freight market data and the customer profile are respectively input into the potential evaluation model; Based on the potential evaluation model, the historical freight market data is used to predict turnover and route profit, and the main evaluation results are obtained. Based on the potential evaluation model, the customer profile is applied to output the revised evaluation result.

[0086] Based on any of the above embodiments, the evaluation unit is further specifically used for: Extract customer profile features from the customer profile; The revised evaluation result is obtained by applying the customer profile features based on the potential evaluation model.

[0087] Based on any of the above embodiments, the acquisition unit is specifically used for: Collect initial historical freight market data; The initial historical freight market data is preprocessed to obtain clean market data; Perform a data verification operation on the cleaning market data to obtain the historical freight market data; The data verification operation includes at least one of data integrity verification, logical consistency verification, and outlier detection.

[0088] Based on any of the above embodiments, the historical freight market data includes at least one of macroeconomic data, industry traffic data, publicly available market data, and road-enterprise interaction data; The customer profile includes at least one of the following: customer industry attributes, frequency of cargo transportation, volume of transportation, payment ability, and historical stability of cooperation.

[0089] Based on any of the above embodiments, an adjustment unit is included after the fusion unit, and the adjustment unit is specifically used for: Based on a supervised learning algorithm, abnormal results in the market potential evaluation results are analyzed. If the abnormal result exceeds the warning threshold, the parameters of the potential evaluation model are adjusted.

[0090] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include a processor 410, a communication interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a method for evaluating the potential of the railway freight market. This method includes: acquiring historical freight market data and customer profiles; inputting the historical freight market data and the customer profiles into a potential evaluation model to obtain a main evaluation result and a revised evaluation result output by the potential evaluation model; and obtaining a market potential evaluation result based on the main evaluation result and the revised evaluation result. The potential evaluation model is constructed based on a neural network.

[0091] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0092] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the railway freight market potential evaluation method provided by the above methods. The method includes: acquiring historical freight market data and customer profiles; inputting the historical freight market data and the customer profiles into a potential evaluation model to obtain a main evaluation result and a revised evaluation result output by the potential evaluation model; and obtaining a market potential evaluation result based on the main evaluation result and the revised evaluation result. The potential evaluation model is constructed based on a neural network.

[0093] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for evaluating the potential of the railway freight market provided by the methods described above. This method includes: acquiring historical freight market data and customer profiles; inputting the historical freight market data and the customer profiles into a potential evaluation model to obtain a primary evaluation result and a revised evaluation result output by the potential evaluation model; and obtaining a market potential evaluation result based on the primary evaluation result and the revised evaluation result. The potential evaluation model is constructed based on a neural network.

[0094] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0095] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0096] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating the market potential of railway freight, characterized in that, include: Obtain historical freight market data and customer profiles; The historical freight market data and the customer profile are respectively input into the potential evaluation model to obtain the main evaluation result and the revised evaluation result output by the potential evaluation model. Based on the main evaluation results and the revised evaluation results, the market potential evaluation results are obtained; The potential evaluation model is constructed based on a neural network.

2. The method for evaluating the potential of the railway freight market according to claim 1, characterized in that, The process involves inputting the historical freight market data and the customer profile into the potential evaluation model, respectively, to obtain the main evaluation result and the revised evaluation result output by the potential evaluation model, including: The historical freight market data and the customer profile are respectively input into the potential evaluation model; Based on the potential evaluation model, the historical freight market data is used to predict turnover and route profit, and the main evaluation results are obtained. Based on the potential evaluation model, the customer profile is applied to output the revised evaluation result.

3. The method for evaluating the potential of the railway freight market according to claim 2, characterized in that, The process of applying the customer profile to the potential evaluation model and outputting the revised evaluation result includes: Extract customer profile features from the customer profile; The revised evaluation result is obtained by applying the customer profile features based on the potential evaluation model.

4. The method for evaluating the potential of the railway freight market according to any one of claims 1 to 3, characterized in that, The steps for obtaining the historical freight market data include: Collect initial historical freight market data; The initial historical freight market data is preprocessed to obtain clean market data; Perform a data verification operation on the cleaning market data to obtain the historical freight market data; The data verification operation includes at least one of data integrity verification, logical consistency verification, and outlier detection.

5. The method for evaluating the potential of the railway freight market according to any one of claims 1 to 3, characterized in that, The historical freight market data includes at least one of the following: macroeconomic data, industry transportation data, publicly available market data, and data exchanged between road and enterprise. The customer profile includes at least one of the following: customer industry attributes, frequency of cargo transportation, volume of transportation, payment ability, and historical stability of cooperation.

6. The method for evaluating the potential of the railway freight market according to any one of claims 1 to 3, characterized in that, The process of obtaining a market potential evaluation result based on the main evaluation result and the revised evaluation result, followed by: Based on a supervised learning algorithm, abnormal results in the market potential evaluation results are analyzed. If the abnormal result exceeds the warning threshold, the parameters of the potential evaluation model are adjusted.

7. A device for evaluating the potential of the railway freight market, characterized in that, include: The acquisition unit retrieves historical freight market data and customer profiles. The evaluation unit inputs the historical freight market data and the customer profile into the potential evaluation model to obtain the main evaluation result and the revised evaluation result output by the potential evaluation model. The fusion unit obtains the market potential evaluation result based on the main evaluation result and the revised evaluation result; The potential evaluation model is constructed based on a neural network.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for evaluating the potential of the railway freight market as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method for evaluating the potential of the railway freight market as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method for evaluating the potential of the railway freight market as described in any one of claims 1 to 6.

Images