A method and system for short-term forecasting of highway freight volume

Abstract

This invention relates to the field of transportation statistical forecasting technology, and discloses a method and system for short-term forecasting of highway freight volume. Addressing the problems of low accuracy and poor scenario adaptability in existing forecasting methods, this method collects and preprocesses multi-source scenario-specific freight data, performs first-order differencing and seasonal differencing stabilization on non-stationary time-series data, introduces adaptive weighting coefficients, constructs an improved seasonal ARIMA model, and uses the weighted AIC criterion to select optimal parameters, outputting the month-on-month growth rate and preliminary forecast values. Finally, it combines the latest measured base period value with the adaptive weighting coefficients for double correction to obtain the final forecast value. This invention can accurately reflect the trend and seasonal fluctuations of highway freight volume, effectively reducing the cumulative error of long-step forecasts. It is suitable for short-term assessment of monthly and quarterly highway freight volume, and is particularly suitable for regional freight forecasting of goods with a high proportion of freight volume and significant seasonal fluctuations. It is highly practical and provides stable and reliable forecasts.

Description

Technical Field

[0001] This invention relates to the field of transportation statistical forecasting technology, specifically to a method and system for short-term forecasting of highway freight volume. Background Technology

[0002] Highway freight volume is a core synchronous indicator reflecting the macroeconomic situation and a key basis for monitoring the operation of the transportation industry, road network scheduling, industry planning, and freight control and management decisions. Its monthly, quarterly, and semi-annual short-term forecasts directly affect the scientific and rational nature of transportation resource allocation, road network traffic control, and freight market supervision. Given that highway freight in northern regions is significantly affected by factors such as the transport of special goods, seasonal climate, holiday shutdowns, and extreme weather, freight volume data exhibits a triple characteristic: long-term non-stationary trends, annual cyclical fluctuations, and short-term random disturbances. Conventional forecasting methods struggle to achieve accurate fitting of these characteristics.

[0003] Existing short-term forecasting techniques for highway freight volume mostly employ single trend extrapolation methods, exponential smoothing methods, traditional regression models, or unoptimized conventional time series models. These methods fail to specifically adapt to the non-stationary characteristics, annual and seasonal fluctuations, and sudden disturbances of highway freight volume. Model parameter selection relies solely on theoretical criteria and lacks verification and screening mechanisms tailored to actual freight scenarios, resulting in low forecast accuracy. Furthermore, most existing methods directly output absolute forecast values ​​without incorporating actual freight volume fluctuation patterns and scenario characteristics for error correction. As the forecast step size increases, the accumulated error gradually amplifies, leading to insufficient stability and reliability of the forecast results. These methods cannot accurately reflect the freight volume changes in regions with a high proportion of coal transportation, significant seasonal fluctuations, and frequent external disturbances, thus failing to meet the actual needs of the transportation industry for refined statistics, precise decision-making, and efficient scheduling.

[0004] In view of the technical shortcomings of existing technologies, such as imperfect processing of non-stationary sequences, non-standard selection of model parameters, poor adaptability to scenarios, large prediction errors, and weak anti-disturbance ability, there is an urgent need for a short-term prediction method that fits the fluctuation characteristics of highway freight volume data, adapts to regional freight scenarios, has high prediction accuracy, strong stability, and excellent anti-interference ability, so as to solve the various shortcomings of existing technologies. Summary of the Invention

[0005] To address the aforementioned shortcomings in existing technologies, this invention provides a method and system for short-term forecasting of highway freight volume. Based on the traditional seasonal ARIMA model, it makes scenario-based improvements by introducing adaptive weighted correction coefficients and scenario-based parameter selection criteria. Addressing the unique characteristics of highway freight, such as seasonal fluctuations, special cargo transportation, and holiday disturbances, it performs stabilization preprocessing on historical freight volume time-series data, constructs an improved seasonal ARIMA model adapted to the freight scenario, and selects the optimal parameters. The preliminary forecast results are then double-corrected by combining month-on-month growth rate and adaptive weighted coefficients. This effectively reflects the trend changes, seasonal fluctuations, and short-term disturbance patterns of highway freight volume, significantly improving the accuracy, stability, and scenario adaptability of short-term forecasts. It solves the technical problems of large errors, poor adaptability, unsatisfactory fitting effects, and weak anti-disturbance capabilities of traditional forecasting methods.

[0006] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows:

[0007] In a first aspect, the present invention proposes a short-term forecasting method for highway freight volume, comprising the following steps:

[0008] Collect historical data of monthly highway freight volume in the target area for several consecutive years and scenario-related indicator data for the same period, and preprocess them to obtain the time series of monthly highway freight volume.

[0009] The stationarity of the monthly highway freight volume time series was tested, and the non-stationary time series were successively subjected to first-order differencing and seasonal differencing with a period of 12 to obtain the stationary difference series of monthly highway freight volume.

[0010] A seasonal ARIMA model was constructed, and the optimal parameter combination of the model was determined by the adaptive weighted AIC criterion based on the stationary difference series of monthly road freight volume, thus obtaining the optimal seasonal ARIMA model.

[0011] The optimal seasonal ARIMA model is used to predict the stationary difference series of monthly highway freight volume, and the short-term month-on-month growth rate of highway freight volume is obtained.

[0012] Based on the most recent monthly highway freight volume historical data, combined with adaptive weighting coefficients and short-term monthly highway freight volume month-on-month growth rate, the short-term monthly highway freight volume forecast is calculated.

[0013] Furthermore, the scenario-based correlation indicator data includes one or more of the following: cargo output, road freight share, macroeconomic indicators, holiday data, and climate data.

[0014] Furthermore, the preprocessing includes: correcting abnormal fluctuation data caused by holidays, extreme weather, or sudden control measures using the seasonal average of the same period; and filling in missing data using linear interpolation.

[0015] Furthermore, the formula for calculating the first-order difference is:

[0016] ;

[0017] in, Let be the first difference result of the road freight volume in month t. Let be the road freight volume for month t. This represents the road freight volume for month t-1.

[0018] The formula for calculating the seasonal difference with a period of 12 is as follows:

[0019] ;

[0020] in, The seasonal difference results are based on a period of 12. Let be the first difference result of the road freight volume in month t. The result is the first difference of the road freight volume for month t-12.

[0021] Furthermore, the constructed seasonal ARIMA model is as follows:

[0022] ;

[0023] in, It is an autoregressive polynomial. For non-seasonal autoregressive order, For delay operators, For seasonal autoregressive polynomials, For the order of seasonal autoregression, The seasonal difference results are based on a period of 12. It is the first-order difference order. Let the seasonal difference order be , Let be the road freight volume for month t. It is a moving average polynomial. The order of the non-seasonal moving average. It is a seasonally sliding average polynomial. The order of the seasonal moving average. This is a white noise sequence for month t.

[0024] Furthermore, the adaptive weighted AIC criterion is calculated as follows:

[0025]

[0026] in, The weighted AIC value of the model. Let be the maximum likelihood function of the model. For adaptive weighting coefficients, This represents the number of model parameters.

[0027] Furthermore, based on the stationary difference series of monthly highway freight volume, the optimal parameter combination of the model is determined using the adaptive weighted AIC criterion, resulting in the optimal seasonal ARIMA model, including:

[0028] Based on the characteristics of the autocorrelation function and partial autocorrelation function of the stationary difference series of monthly highway freight volume, the range of candidate parameter values ​​for non-seasonal autoregression order, non-seasonal moving average order, seasonal autoregression order, and seasonal moving average order are set.

[0029] Construct seasonal ARIMA models for each candidate parameter combination;

[0030] The weighted AIC value of the seasonal ARIMA model corresponding to each candidate parameter combination is calculated according to the adaptive weighted AIC criterion, and the model with the smallest weighted AIC value is selected as the initial optimal seasonal ARIMA model.

[0031] The initial optimal seasonal ARIMA model is subjected to residual white noise test and goodness-of-fit evaluation. If the test is passed, the model is confirmed as the optimal seasonal ARIMA model. Otherwise, the model that passes the test is selected as the optimal seasonal ARIMA model from the candidate models with smaller weighted AIC values.

[0032] Furthermore, the calculation method for short-term monthly road freight volume forecasts is as follows:

[0033]

[0034] in, This represents the short-term monthly road freight volume forecast for month t. For adaptive weighting coefficients, This is historical data on the most recent monthly road freight volume. This represents the month-on-month growth rate of short-term highway freight volume.

[0035] Furthermore, it also includes:

[0036] The corresponding short-term monthly road freight turnover forecast is calculated by multiplying the short-term monthly road freight volume forecast by the average road freight distance in the target area over the past year.

[0037] Secondly, this invention also proposes a short-term forecasting system for highway freight volume, which applies the aforementioned short-term forecasting method for highway freight volume, including:

[0038] The data acquisition and preprocessing module is used to collect historical data of monthly highway freight volume in the target area over several consecutive years, as well as scenario-related indicator data for the same period, and preprocess them to obtain the time series of monthly highway freight volume.

[0039] The sequence stationarization module is used to perform stationarity tests on the monthly highway freight volume time series, and to perform first-order differencing and seasonal differencing with a period of 12 on the non-stationary time series to obtain the stationary difference series of monthly highway freight volume.

[0040] The model building and training module is used to build a seasonal ARIMA model and, based on the stationary difference series of monthly road freight volume, uses the adaptive weighted AIC criterion to determine the optimal combination of parameters for the model, thus obtaining the optimal seasonal ARIMA model.

[0041] The month-on-month growth rate prediction module is used to predict the monthly road freight volume stationary difference series using the optimal seasonal ARIMA model, and obtain the short-term monthly road freight volume month-on-month growth rate.

[0042] The freight volume forecasting module is used to calculate the short-term monthly freight volume forecast based on the most recent monthly historical data of highway freight volume, combined with an adaptive weighting coefficient and the short-term monthly growth rate of highway freight volume.

[0043] The present invention has the following beneficial effects:

[0044] (1) This invention addresses the complex characteristics of road freight volume, which has long-term trends, annual seasonal fluctuations, and short-term random disturbances. By using first-order difference and periodic seasonal difference dual stabilization processing, non-stationary time series data is transformed into stationary sequences, breaking through the technical barrier that conventional models cannot adapt to non-stationary freight data and greatly improving the model's adaptability to different scenarios.

[0045] (2) This invention improves the traditional seasonal ARIMA model by introducing adaptive weighting coefficients and weighting information criteria, abandoning the pure theoretical parameter selection logic, and dynamically optimizing parameters in combination with the fluctuation range of road freight, avoiding the arbitrariness of parameter selection, resulting in a higher model fit, which can accurately fit the regional road freight volume change pattern dominated by special goods transportation, significant seasonal fluctuations, and frequent external disturbances, and has stronger anti-interference ability.

[0046] (3) This invention addresses the practical needs of short-term forecasting of highway freight. It abandons the traditional approach of directly using the absolute forecast value of the model. Instead, it relies on the model's accurate month-on-month growth rate forecast and combines recent measured base period data and adaptive coefficients to double-correct the final forecast value. This effectively weakens the cumulative error of long-step forecasts and specifically offsets scenario-based disturbance errors such as holidays and climate. It significantly improves the accuracy and reliability of short-term forecast results and fully meets the needs of accurate monthly, quarterly, and semi-annual assessments, industry decision-making, and resource scheduling of highway freight volume.

[0047] (4) This invention incorporates multi-source scenario-based feature data. In the data preprocessing stage, it uses holiday data, climate data, etc., to help locate abnormal fluctuation points. Outliers caused by Spring Festival shutdowns, extreme weather, etc., are corrected for the seasonal mean of the same period, rather than simply being removed, thus preserving the true freight data patterns to the greatest extent. In the model parameter selection stage, the adaptive weighting coefficient is dynamically adjusted according to the seasonal fluctuation characteristics of indicators such as cargo output and road freight share, making the model more in line with the actual scenario. This method balances prediction accuracy and practicality, is more applicable, and is suitable for various short-term road freight volume prediction scenarios. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of a short-term forecasting method for highway freight volume according to the present invention.

[0049] Figure 2 This is a schematic diagram of the data stationarity test and the parameter selection process for the optimal seasonal ARIMA improved model in this invention;

[0050] Figure 3 This is a schematic diagram of the structure of a short-term forecasting system for highway freight volume according to the present invention. Detailed Implementation

[0051] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0052] like Figure 1 and Figure 2 As shown, an embodiment of the present invention provides a short-term forecasting method for highway freight volume, comprising the following steps S1 to S5:

[0053] S1. Collect historical data of monthly highway freight volume in the target area for several consecutive years and the corresponding scenario-related indicator data, and preprocess them to obtain the time series of monthly highway freight volume.

[0054] In an optional embodiment of the present invention, the scenario-based correlation index data collected in step S1 includes one or more of the following: cargo output, road freight share, macroeconomic indicators, holiday data, and climate data. This embodiment collects multi-source scenario-based road freight basic data by collecting historical statistical data on monthly road freight volume and freight turnover in the target area over several consecutive years, and simultaneously collects scenario-based correlation indicators such as cargo output, road freight share, macroeconomic indicators, holiday calendar, and temperature and climate data for the same period.

[0055] The collected scenario-based correlation index data is used for two purposes: First, to help identify abnormal fluctuation points. For example, by using holiday data, we can locate the sharp drop in freight volume caused by statutory holidays such as the Spring Festival and National Day, and by using climate data to identify extreme weather windows such as blizzards and rainstorms, so as to correct the abnormal deviation values ​​of the corresponding months using the seasonal average of the same period. Second, in the subsequent model parameter screening stage, we can dynamically adjust the value of the adaptive weighting coefficient according to the seasonal fluctuation characteristics of cargo output (such as the peak season for outbound transportation).

[0056] Step S1, preprocessing, includes: correcting abnormal fluctuations in data caused by holidays, extreme weather, or sudden control measures using the seasonal average of the same period; and filling in missing data using linear interpolation. In this embodiment, the collected raw data is cleaned by using the seasonal average correction method for abnormal fluctuations in freight volume caused by Spring Festival shutdowns, extreme weather, or sudden control measures, rather than directly removing them, thus preserving the true characteristics of freight fluctuations. Linear interpolation is used to fill in discontinuous and missing data, forming a continuous, complete, and standardized monthly time series dataset. The original freight volume time series is denoted as... , where t represents the month of the time.

[0057] In this embodiment, taking freight volume forecasting for a certain region as an example, official statistics on monthly road freight volume and freight turnover for the target region from January 2021 to July 2025 are obtained. The data comes from statistical reports of relevant transportation departments. A total of 55 monthly sample data were obtained during this period. Simultaneously, scenario-related indicators such as the output of transported goods, road freight share, regional GDP, holiday calendar, and temperature and climate are collected.

[0058] The collected raw data is cleaned: For abnormal fluctuations in freight volume in February and October of each year, caused by the Spring Festival holiday, road closures due to blizzards, and sudden control measures, the average monthly data of the same period in the past three years is used for correction instead of direct removal, so as to retain the true characteristics of freight fluctuations; linear interpolation is used to supplement a small number of discontinuous and missing data to form a continuous, complete, and standardized monthly time series dataset.

[0059] S2. The stationarity of the monthly highway freight volume time series is tested, and the non-stationary time series are successively subjected to first-order differencing and seasonal differencing with a period of 12 to obtain the stationary difference series of monthly highway freight volume.

[0060] In an optional embodiment of the present invention, step S2 uses the ADF unit root test combined with serial autocorrelation and partial autocorrelation analysis to determine the original freight volume time series. To determine the stationarity of the series, if the p-value is greater than 0.05, the series is considered non-stationary. First, the original series is subjected to first-order differencing to eliminate the long-term upward or downward trend in freight volume. The formula for calculating the first-order differencing is:

[0061] ;

[0062] in, Let be the first difference result of the road freight volume in month t. Let be the road freight volume for month t. This represents the road freight volume for month t-1.

[0063] Then, the first-order difference sequence is subjected to seasonal differencing with a 12-month cycle to eliminate annual cyclical fluctuations caused by factors such as winter heating, summer production, and peak seasons for the transportation of special goods (coal). The formula for calculating the seasonal differencing is:

[0064] ;

[0065] in, The seasonal difference results are based on a period of 12. Let be the first difference result of the road freight volume in month t. The result is the first difference of the road freight volume for month t-12.

[0066] After double difference calculus, a stationary difference sequence is obtained. Then, perform the ADF unit root test again to ensure that the p-value of the differencing sequence is less than 0.05, which meets the stationarity modeling standard and is suitable for subsequent model fitting requirements.

[0067] S3. Construct a seasonal ARIMA model, and based on the stationary difference series of monthly road freight volume, use the adaptive weighted AIC criterion to determine the optimal parameter combination of the model to obtain the optimal seasonal ARIMA model.

[0068] In an optional embodiment of the present invention, step S3 optimizes the model structure for the short-term fluctuation characteristics of road freight transport and constructs a seasonal ARIMA(p, d, q)(P, D, Q). 12 The improved model is as follows:

[0069] ;

[0070] in, It is an autoregressive polynomial. For non-seasonal autoregressive order, For delay operators, For seasonal autoregressive polynomials, For the order of seasonal autoregression, The seasonal difference results are based on a period of 12. It is the first-order difference order. Let the seasonal difference order be , Let be the road freight volume for month t. It is a moving average polynomial. The order of the non-seasonal moving average. It is a seasonally sliding average polynomial. The order of the seasonal moving average. The sequence is a white noise sequence for month t. Based on the results of step S2, the first-order difference order d=1 and the seasonal difference order D=1 are set.

[0071] In an optional embodiment of the present invention, step S3, based on the stationary difference series of monthly highway freight volume, uses the adaptive weighted AIC criterion to determine the optimal parameter combination of the model, thereby obtaining the optimal seasonal ARIMA model, including:

[0072] Based on the characteristics of the autocorrelation function and partial autocorrelation function of the stationary difference series of monthly highway freight volume, the range of candidate parameter values ​​for non-seasonal autoregression order, non-seasonal moving average order, seasonal autoregression order, and seasonal moving average order are set.

[0073] Construct seasonal ARIMA models for each candidate parameter combination;

[0074] The weighted AIC value of the seasonal ARIMA model corresponding to each candidate parameter combination is calculated according to the adaptive weighted AIC criterion, and the model with the smallest weighted AIC value is selected as the initial optimal seasonal ARIMA model.

[0075] The initial optimal seasonal ARIMA model is subjected to residual white noise test and goodness-of-fit evaluation. If the test is passed, the model is confirmed as the optimal seasonal ARIMA model. Otherwise, the model that passes the test is selected as the optimal seasonal ARIMA model from the candidate models with smaller weighted AIC values.

[0076] This embodiment analyzes the autocorrelation function (ACF) and partial autocorrelation function (PACF) characteristics of stationary differencing sequences. At non-seasonal lags (1, 2, 3…), both ACF and PACF exhibit tailing characteristics after first-order differencing, indicating that the non-seasonal portion should simultaneously include autoregressive and moving average terms. Therefore, the candidate range for the non-seasonal autoregressive order p is set to {0, 1, 2}, and the candidate range for the non-seasonal moving average order q is set to {0, 1}. At seasonal lags (12, 24…), the ACF and PACF after seasonal differencing exhibit truncation characteristics at lag 12, indicating that the seasonal portion only requires a moving average term. Therefore, the candidate range for the seasonal autoregressive order P is set to {0}, and the candidate range for the seasonal moving average order Q is set to {0, 1}. A total of (3 × 2 × 1 × 2) = 12 candidate parameter combinations are obtained, and each candidate seasonal ARIMA model is constructed for subsequent screening.

[0077] Meanwhile, an adaptive weighted AIC criterion tailored to the freight scenario is introduced to select the optimal parameters. The calculation method of the adaptive weighted AIC criterion is as follows:

[0078]

[0079] in, The weighted AIC value of the model. Let be the maximum likelihood function of the model. For adaptive weighting coefficients, This represents the number of model parameters. This embodiment introduces adaptive weighting coefficients. To optimize model fit, the adaptive weighting coefficient typically ranges from 0.85 to 1.15. This is particularly important for months with significant freight fluctuations, such as peak seasons for outbound freight and periods of high demand during winter. The value is typically between 1.05 and 1.15, targeting periods of low freight volume. The value is usually between 0.85 and 0.95, which corresponds to the actual fluctuation range.

[0080] This embodiment, considering the characteristics of freight fluctuations, sets the baseline value of the adaptive weighting coefficient λ to 1.0 (using standard weighting). The WAIC values ​​of all candidate models are calculated, and the model with the smallest WAIC value is selected as the initial optimal model. Based on this, auxiliary validation of the initial optimal model is performed: its goodness of fit (R²) is calculated, and a residual white noise test is conducted. The validation shows that the model has a goodness of fit of 0.87, and the residual white noise test p-value > 0.05, indicating that the model has fully extracted sequence information and the residuals have no autocorrelation. Simultaneously, it is compared with other models with similar WAIC values ​​(such as ARIMA(2,1,0)(0,1,1)). 12 (The WAIC value is slightly higher than 0.3), confirming the selected ARIMA(1,1,1)(0,1,1) 12The overall performance was optimal. The final optimal model was determined to be ARIMA(1,1,1)(0,1,1). 12 Complete the model parameter estimation and significance verification to ensure that the model fits the regional freight transport patterns.

[0081] Therefore, the computable prediction equation is: .

[0082] S4. The optimal seasonal ARIMA model is used to predict the stationary difference series of monthly highway freight volume to obtain the short-term month-on-month growth rate of highway freight volume.

[0083] In an optional embodiment of the present invention, step S4 substitutes the optimal seasonal ARIMA model into the stationary difference series to perform short-term monthly and quarterly forecasts, thereby obtaining the month-on-month growth rate of freight volume. Compared with the preliminary absolute forecast value By performing error analysis on the model prediction results, it was determined that the model can accurately fit the long-term trend and seasonal fluctuation pattern of freight volume. However, the directly output absolute prediction value will generate a small cumulative error as the prediction time increases. Therefore, the month-on-month growth rate data output by the model is retained, and the subsequent scenario-based results are corrected by combining the adaptive coefficient.

[0084] This embodiment substitutes the optimal seasonal ARIMA model into a stationary difference series to make short-term forecasts for September to December 2025, obtaining the month-on-month growth rate of freight volume for each forecast month. and preliminary absolute forecast values The model initially predicts a total freight volume of 411.5 million tons for September-December 2025. Error analysis of the prediction results shows that the model can accurately fit the trend and seasonal fluctuations of freight volume. The predicted curve is largely consistent with historical seasonal patterns in terms of direction and phased trends, and can reproduce the typical road freight evolution pattern of "September remaining at a relative peak, with an overall decline in the fourth quarter." However, the directly output absolute prediction value has a slight cumulative error as the prediction period increases. Compared with the actual total of 481.2 million tons for September-December 2025, the initial prediction is 14.5% lower. Therefore, the month-on-month growth rate data output by the model is retained for subsequent scenario-based result correction.

[0085] S5. Based on the most recent monthly highway freight volume historical data, combined with the adaptive weighting coefficient and the short-term monthly highway freight volume month-on-month growth rate, the short-term monthly highway freight volume forecast value is calculated.

[0086] In an optional embodiment of the present invention, step S5 selects the latest officially measured monthly highway freight volume data based on the month-on-month growth rate of each predicted month output by the optimal model. As a benchmark value, an adaptive weighting coefficient is introduced. Contextual corrections are performed to offset short-term random disturbance errors. The correction formula is as follows:

[0087]

[0088] in, This represents the short-term monthly road freight volume forecast for month t. For adaptive weighting coefficients, This is historical data on the most recent monthly road freight volume. This represents the month-on-month growth rate of short-term highway freight volume.

[0089] In this embodiment, the month-on-month growth rate for each predicted month is based on the output of the optimal model. The official measured monthly road freight volume of 126.8 million tons in August 2025 was selected as the benchmark value. Adaptive weighting coefficients are set based on the characteristics of the predicted month. September 2025 is the transitional peak season for freight transport from summer to autumn. Freight demand will decline but will be slightly better than model expectations. The figure was 1.04; October is traditionally a slow season, so the actual decline was slightly less than predicted. Take 1.02; November is a stable period. The actual decrease was 1.00; December is the off-season for freight transport during the winter year-end settlement period, and the actual decrease was significantly smaller than predicted. We take 1.08. Substituting the base period measured value, the adaptive coefficient, and the model-predicted month-on-month growth rate for the corresponding month into the correction formula, we obtain the corrected forecast value for highway freight volume. The results show that the revised prediction error for September converged from -7.2% initially predicted by the model to -1.1%, and the error for December converged from -22.6% to -2.5%, with the total error converging from -14.5% to -2.2%, indicating a significant improvement in prediction accuracy. The amplified error in December was mainly due to structural disturbances in the fourth quarter (such as a sudden drop in temperature inhibiting construction and staggered production), which the linear seasonality and variation could not fully internalize. However, the overall trend pattern remained basically correct.

[0090] In an optional embodiment of the present invention, the present invention further includes:

[0091] Based on short-term monthly road freight volume forecasts Average road freight distance to the target area over the past year The product of these factors is used to calculate the corresponding short-term monthly highway freight turnover forecast. The calculation formula is:

[0092] ;

[0093] Complete short-term forecasting of the entire process of highway freight transport volume.

[0094] This embodiment calculates the average road freight distance for the province over the past year to be 168 kilometers. By multiplying the corrected freight volume by the average distance, the corresponding predicted freight turnover is obtained. The calculation formula is:

[0095] ;

[0096] The results show that the overall deviation direction of the turnover volume forecast is basically consistent with that of the freight volume forecast. The forecast errors for September and October are relatively small (-1.1% and -1.7%, respectively), while the forecast errors for November and December are relatively high. Overall, the total forecast error is within a controllable range. This indicates that when the forecast is linearly transmitted using "average transport distance × freight volume," a higher error in the freight volume forecast will proportionally amplify the deviation in the turnover volume forecast.

[0097] This embodiment verifies the effectiveness and practicality of the short-term forecasting method for highway freight volume proposed in this invention. By introducing adaptive weighting coefficients and a dual correction strategy, this method can effectively reduce the cumulative error of long-step forecasting and improve forecasting accuracy, and is particularly suitable for regional freight forecasting with a high proportion of special goods (coal) transportation and significant seasonal fluctuations.

[0098] like Figure 3 As shown, this embodiment of the invention also provides a short-term forecasting system for highway freight volume, which applies the short-term forecasting method for highway freight volume described above, including:

[0099] The data acquisition and preprocessing module is used to collect historical data of monthly highway freight volume in the target area over several consecutive years, as well as scenario-related indicator data for the same period, and preprocess them to obtain the time series of monthly highway freight volume.

[0100] The sequence stationarization module is used to perform stationarity tests on the monthly highway freight volume time series, and to perform first-order differencing and seasonal differencing with a period of 12 on the non-stationary time series to obtain the stationary difference series of monthly highway freight volume.

[0101] The model building and training module is used to build a seasonal ARIMA model and, based on the stationary difference series of monthly road freight volume, uses the adaptive weighted AIC criterion to determine the optimal combination of parameters for the model, thus obtaining the optimal seasonal ARIMA model.

[0102] The month-on-month growth rate prediction module is used to predict the monthly road freight volume stationary difference series using the optimal seasonal ARIMA model, and obtain the short-term monthly road freight volume month-on-month growth rate.

[0103] The freight volume forecasting module is used to calculate the short-term monthly freight volume forecast based on the most recent monthly historical data of highway freight volume, combined with an adaptive weighting coefficient and the short-term monthly growth rate of highway freight volume.

[0104] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0105] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0106] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0107] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

[0108] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A method for short-term forecasting of highway freight volume, characterized in that, Includes the following steps: Collect historical data of monthly highway freight volume in the target area for several consecutive years and scenario-related indicator data for the same period, and preprocess them to obtain the time series of monthly highway freight volume. The stationarity of the monthly highway freight volume time series was tested, and the non-stationary time series were successively subjected to first-order differencing and seasonal differencing with a period of 12 to obtain the stationary difference series of monthly highway freight volume. A seasonal ARIMA model is constructed, and based on the stationary difference series of monthly highway freight volume, the optimal parameter combination of the model is determined using the adaptive weighted AIC criterion, thus obtaining the optimal seasonal ARIMA model. The calculation method of the adaptive weighted AIC criterion is as follows: ； in, The weighted AIC value of the model. Let be the maximum likelihood function of the model. For adaptive weighting coefficients, The number of model parameters; The optimal seasonal ARIMA model is used to predict the stationary difference series of monthly highway freight volume, and the short-term month-on-month growth rate of highway freight volume is obtained. Based on the most recent historical monthly highway freight volume data, combined with adaptive weighting coefficients and the short-term monthly highway freight volume month-on-month growth rate, the short-term monthly highway freight volume forecast is calculated as follows: ； in, This represents the short-term monthly road freight volume forecast for month t. For adaptive weighting coefficients, This is historical data on the most recent monthly road freight volume. This represents the month-on-month growth rate of short-term highway freight volume.

2. The method for short-term forecasting of highway freight volume according to claim 1, characterized in that, The scenario-based related indicator data includes one or more of the following: cargo output, road freight share, macroeconomic indicators, holiday data, and climate data.

3. The method for short-term forecasting of highway freight volume according to claim 1, characterized in that, The preprocessing includes: correcting abnormal fluctuations in data caused by holidays, extreme weather, or sudden control measures using the seasonal average of the same period; and filling in missing data using linear interpolation.

4. The method for short-term forecasting of highway freight volume according to claim 1, characterized in that, The formula for calculating the first-order difference is: ； in, Let be the first difference result of the road freight volume in month t. Let be the road freight volume for month t. This represents the road freight volume for month t-1. The formula for calculating the seasonal difference with a period of 12 is as follows: ； in, The seasonal difference results are based on a period of 12. Let be the first difference result of the road freight volume in month t. The result is the first difference of the road freight volume for month t-12.

5. The method for short-term forecasting of highway freight volume according to claim 1, characterized in that, The constructed seasonal ARIMA model is as follows: ； in, It is an autoregressive polynomial. For non-seasonal autoregressive order, For delay operators, For seasonal autoregressive polynomials, For the order of seasonal autoregression, The seasonal difference results are based on a period of 12. It is the first-order difference order. Let the seasonal difference order be , Let be the road freight volume for month t. It is a moving average polynomial. The order of the non-seasonal moving average. It is a seasonally sliding average polynomial. The order of the seasonal moving average. This is a white noise sequence for month t.

6. The method for short-term forecasting of highway freight volume according to claim 1, characterized in that, Based on the stationary difference series of monthly highway freight volume, the optimal parameter combination of the model is determined using the adaptive weighted AIC criterion, resulting in the optimal seasonal ARIMA model, including: Based on the characteristics of the autocorrelation function and partial autocorrelation function of the stationary difference series of monthly highway freight volume, the range of candidate parameter values ​​for non-seasonal autoregression order, non-seasonal moving average order, seasonal autoregression order, and seasonal moving average order are set. Construct seasonal ARIMA models for each candidate parameter combination; The weighted AIC value of the seasonal ARIMA model corresponding to each candidate parameter combination is calculated according to the adaptive weighted AIC criterion, and the model with the smallest weighted AIC value is selected as the initial optimal seasonal ARIMA model. The initial optimal seasonal ARIMA model is subjected to residual white noise test and goodness-of-fit evaluation. If the test is passed, the model is confirmed as the optimal seasonal ARIMA model. Otherwise, the model that passes the test is selected as the optimal seasonal ARIMA model from the candidate models with smaller weighted AIC values.

7. The method for short-term forecasting of highway freight volume according to claim 1, characterized in that, Also includes: The corresponding short-term monthly road freight turnover forecast is calculated by multiplying the short-term monthly road freight volume forecast by the average road freight distance in the target area over the past year.

8. A short-term forecasting system for highway freight volume, employing a short-term forecasting method for highway freight volume as described in any one of claims 1 to 7, characterized in that, include: The data acquisition and preprocessing module is used to collect historical data of monthly highway freight volume in the target area over several consecutive years, as well as scenario-related indicator data for the same period, and preprocess them to obtain the time series of monthly highway freight volume. The sequence stationarization module is used to perform stationarity tests on the monthly highway freight volume time series, and to perform first-order differencing and seasonal differencing with a period of 12 on the non-stationary time series to obtain the stationary difference series of monthly highway freight volume. The model building and training module is used to build a seasonal ARIMA model and, based on the stationary difference series of monthly road freight volume, uses the adaptive weighted AIC criterion to determine the optimal combination of parameters for the model, thus obtaining the optimal seasonal ARIMA model. The month-on-month growth rate prediction module is used to predict the stationary difference series of monthly road freight volume using the optimal seasonal ARIMA model, and obtain the short-term month-on-month growth rate of road freight volume. The freight volume forecasting module is used to calculate the short-term monthly freight volume forecast based on the most recent monthly historical data of highway freight volume, combined with an adaptive weighting coefficient and the short-term monthly growth rate of highway freight volume.

Images