A method and a system for dynamic freight price estimation

Abstract

The invention relates a method which operates on a computer-aided system (S) containing at least one processor and at least one memory unit, is designed to enable rapid and dynamic pricing between freight carriers and freight owners in road transportation characterized by; separated the shipments based on the pricing per quantity or per ton; the data related to the shipment where the transportation process is carried out is predicted independently from each other by at least two artificial intelligence algorithms under the control of the mentioned processor; determining the artificial intelligence algorithm with the lowest error rate among the results of the mentioned artificial intelligence algorithms under the control of at least one processor and generating results using the outcome of the best-performing algorithm; taking the average with considering standard deviation; obtaining correction data by filtering the prices from past shipments based on the characteristics of that shipment, with a higher weight given to prices from the nearest dates shipments at the time of creating the shipment; obtaining the related prediction result by multiplying the data results from the previously selected shipment with the result of the selected artificial intelligence algorithm and adding it to the result of the correction data algorithm.

Description

[0001] A METHOD AND A SYSTEM FOR DYNAMIC FREIGHT PRICE ESTIMATION

[0002] FIELD OF THE INVENTION

[0003] The invention is related to a method and a system that enables the fast and dynamic prediction of pricing between freight carriers and freight owners in road transportation and presents it to users.

[0004] PRIOR ART

[0005] Transportation activities, which are central to supply chains, are of critical importance when considering both domestic and international operations. Efficiency and performance requirements are increasingly growing in road transport operations. At this point, innovations required by technology to enhance efficiency in operations are becoming more widespread every day. Due to the extensive road network and suitable transportation vehicles available worldwide, road transport, which provides the opportunity for Full Truck Load (FTL) transportation, is an important research area that deserves attention and digitalization from a commercial perspective. Dynamic freight pricing in transportation operations is one of the major challenges.

[0006] In road transportation, digital platform structures are used to facilitate interaction between freight carriers and freight owners, bringing together those in need of transportation services with those who transport goods. Especially in freight transportation, the processes of finding freight and finding a suitable carrier can be economically and reliably achieved through these mentioned digital platforms.

[0007] Parameters directly affecting the cost in transportation can continuously vary depending on the region / route where the transportation process takes place. (For example, factors like fluctuating fuel prices based on location, the maximum time the carrier will be on the road, variable road conditions, and transportation times influenced by climate.) In this context, generating dynamic prices for routes through digital medium plays an important role in minimizing transportation costs. Since fluctuating fuel prices and financial parameters directly affect route-based pricing, it is important to pre-determine the current freight rate based on these parameters. There is a need to transparently connect the current and appropriately estimated shipments with both customers and drivers through a digital platform. In the existing technology, there is no method or solution on digital platforms that can accurately predict freight rates, transmit these predictions to operational units, and increase the buy-sell margin through reducing current freight costs in the marketplace.

[0008] The USPTO patent document with publication number US20220318746A1 discloses a system that manages transportation using dynamic routes. In this system, parameters are gathered for moving a load from an origin point to a destination point. Strategic capacity profiles containing meta-tags are used to determine a strategy for carrying the freight, with the strategy being defined by meta-tags that describe carrier attributes used for selecting a carrier. These meta-tags are utilized to define a set of carriers that are suitable for carrying the load based on the attributes corresponding to the meta-tags and their data is presented to the user to allow for one or more carrier selections. However, this digital platform does not include a solution for predicting variable costs in advance.

[0009] The SIPO patent document with publication number CN1 15345557A discloses a computer-based pricing method. Within this solution, data is predicted using predetermined estimation coefficients. However, as you mentioned, the invention does not provide the necessary solution for freight pricing.

[0010] As a result, all abovementioned problems have made it necessary to make an improvement in the relevant technical field.

[0011] AIM OF THE INVENTION

[0012] The present invention aims to eliminate the abovementioned problems and to make a development in the relevant technical field.

[0013] The main objective of the invention is to establish a method structure suitable for operation in a computer system within a digital transportation platform that brings together freight carriers and freight owners; that dynamically predict specific freightbased pricing for transportation and provide accurate pricing to the user. Another objective of the invention is to create a mathematical model alongside traditional prediction methods to ensure more accurate estimation with minimal error.

[0014] Another objective of the invention is to ensure healthy trade between two users who receive and provide services by predicting the price in advance.

[0015] Another objective of the invention is to make transportation more economical by allowing freight owners to determine a more accurate transportation fee.

[0016] Another objective of the invention is to eliminate price variations due to unforeseen conditions during transportation.

[0017] Another objective of the invention is introducing a method structure that reduces the workload of companies and enables faster response to requests.

[0018] Another objective of the invention is to provide a price estimate that is minimally affected by sudden inflation and foreign exchange changes.

[0019] BRIEF DESCRIPTION OF THE INVENTION

[0020] The invention is related to a method and a system for dynamic freight price estimation, so as to fulfil all aims mentioned above and will be obtained from the following detailed description.

[0021] The invention relates a method which operates on a computer-aided system containing at least one processor and at least one memory unit, is designed to enable rapid and dynamic pricing between freight carriers and freight owners in road transportation characterized by; separated the shipments based on the pricing per quantity or per ton; the data related to the shipment where the transportation process is carried out is predicted independently from each other by at least two artificial intelligence algorithms under the control of the mentioned processor; determining the artificial intelligence algorithm with the lowest error rate among the results of the mentioned artificial intelligence algorithms under the control of at least one processor and generating results using the outcome of the best-performing algorithm; taking the average with considering standard deviation; obtaining correction data by filtering the prices from past shipments based on the characteristics of that shipment, with a higher weight given to prices from the nearest dates shipments at the time of creating the shipment; obtaining the related prediction result by multiplying the data results from the previously selected shipment with the result of the selected artificial intelligence algorithm and adding it to the result of the correction data algorithm.

[0022] The invention relates a computer-aided system that containing at least one processor and at least one memory unit, is designed to enable rapid and dynamic pricing between freight carriers and freight owners in road transportation characterized in that; includes by a shipment separator unit which separates the shipments based on the pricing per quantity or per ton; an artificial intelligence unit containing at least two artificial intelligence modules which enables independent prediction of data related to the shipment where the process is carried out; at least one combinationer unit that determines the best-performing result from the mentioned two artificial intelligence modules and transmits the outcome; a calibrator unit that obtains correction data based on averages with taking into account standard deviations with higher weight to the prices from the nearest dates completed shipments by filtering historical shipment data according to the characteristics of that shipment which created at the time of the shipment; a pruning module that is capable of disabling the mentioned calibrator unit to prevent the inclusion of pricing parameters in the general prediction in the event that a shipment has taken place outside of the pricing related to the route; the decider unit that determines the prediction result by reducing the error rate of the results from the calibrator unit and the combinationer unit .

[0023] In another preferred embodiment of the invention, it characterized by; pruning shipments that falling outside of the average +2 standard deviations of the general route with blocking the parameters to prevent their inclusion in the general prediction; in cases where a shipment has taken place outside of the pricing for the route which are specified with taking the average with considering standard deviation; obtaining correction data by filtering the prices from past shipments based on the characteristics of that shipment, with a higher weight given to prices from the nearest dates shipments at the time of creating the shipment.

[0024] In another preferred embodiment of the invention, it characterized by; determination of the mentioned prediction result steps that is establish by multiplying the data results from the previously selected shipment with the result of the selected artificial intelligence algorithm and adding it to the result of the correction data algorithm with equation.

[0025] In another preferred embodiment of the invention, it characterized by; using at least two of the algorithms which from LightGBM, ExtraTrees, XGBoost, and Random Forest. In the preferred embodiment, all four algorithms are used independently from each other to generate results.

[0026] In another preferred embodiment of the invention, it characterized by; carrying out a pruning to ensure gradual renewal of filters, thereby preventing the consideration of missing data; in scenarios where the number of shipments is less than 3.

[0027] The invention is a method that incorporates machine learning techniques and mathematical models with the goal of establishing a digital learning system for freight prediction. The purpose of the mathematical model is to support and mitigate potential erroneous results from artificial intelligence by using the realized prices of similar shipments on the same route. The extent to which the mathematical model adjusts the artificial intelligence result is determined by the developed decision-making algorithm (model).

[0028] The protection scope of the invention is specified in the claims and cannot be limited to the description made for illustrative purposes in this brief and detailed description. It is clear that a person skilled in the art can present similar embodiments in the light of the above descriptions without departing from the main theme of the invention.

[0029] BRIEF DESCRIPTION OF DRAWINGS

[0030] Figure 1 provides a flowchart illustrating the operation of the invention's method.

[0031] Figure 2 provides a flowchart describing the estimation method.

[0032] Figure 3 provides a flowchart explaining the operation of the pruning unit during estimation.

[0033] Figure 4 provides a representative drawing illustrating the invention's system. The drawings do not necessarily need to be scaled and some details that are not necessary to understand the present invention may have been omitted. In addition, elements that are at least largely identical or at least largely have identical functions are indicated with the same number.

[0034] DESCRIPTION OF THE REFERENCES IN FIGURES

[0035] 10. Shipment separator unit

[0036] 20. Artificial intelligence unit

[0037] 21. Artificial intelligence module

[0038] 30. Combinationer unit

[0039] 40. Calibrator unit

[0040] 41 . Pruning module

[0041] 50. Decider unit

[0042] S. System

[0043] DETAILED DESCRIPTION OF THE INVENTION

[0044] In this detailed description, the invention a method and a system (S) for dynamic freight price estimation is described by means of examples only for clarifying the subject matter such that no limiting effect is created.

[0045] The invention is related to a method and a system (S) that enables the fast and dynamic prediction of pricing between freight carriers and freight owners in road transportation and presents it to users.

[0046] The invention relates a method which operates on a computer-aided system (S) containing at least one processor and at least one memory unit, is designed to enable rapid and dynamic pricing between freight carriers and freight owners in road transportation characterized by; separated the shipments based on the pricing per quantity or per ton; the data related to the shipment where the transportation process is carried out is predicted independently from each other by at least two artificial intelligence algorithms under the control of the mentioned processor; determining the artificial intelligence algorithm with the lowest error rate among the results of the mentioned artificial intelligence algorithms under the control of at least one processor and generating results using the outcome of the best-performing algorithm; taking the average with considering standard deviation; obtaining correction data by filtering the prices from past shipments based on the characteristics of that shipment, with a higher weight given to prices from the nearest dates shipments at the time of creating the shipment; obtaining the related prediction result by multiplying the data results from the previously selected shipment with the result of the selected artificial intelligence algorithm and adding it to the result of the correction data algorithm.

[0047] The invention relates a computer-aided system (S) that containing at least one processor and at least one memory unit, is designed to enable rapid and dynamic pricing between freight carriers and freight owners in road transportation characterized in that; includes by a shipment separator unit (10) which separates the shipments based on the pricing per quantity or per ton; an artificial intelligence unit (20) containing at least two artificial intelligence modules (21 ) which enables independent prediction of data related to the shipment where the process is carried out; at least one combinationer unit (30) that determines the best-performing result from the mentioned two artificial intelligence modules (21 ) and transmits the outcome; a calibrator unit (40) that obtains correction data based on averages with taking into account standard deviations with higher weight to the prices from the nearest dates completed shipments by filtering historical shipment data according to the characteristics of that shipment which created at the time of the shipment; a pruning module (41 ) that is capable of disabling the mentioned calibrator unit (40) to prevent the inclusion of pricing parameters in the general prediction in the event that a shipment has taken place outside of the pricing related to the route; the decider unit (50) that determines the prediction result by reducing the error rate of the results from the calibrator unit (40) and the combinationer unit (30).

[0048] The invention operates on at least one computer-aided system that includes at least one processor and memory unit. Preferably, it operates on at least one server and contains shipment / shipping data to be processed. The mentioned shipment data can be recorded in a computer database by at least one person, or it can be automatically retrieved from a data source through integration. (For example, daily fuel prices, daily exchange rates, etc.) Figure 1 provides a flowchart describing the operation of the invention. Within the scope of the invention, data related to shipments is separated by the shipment separator unit (10), and prediction for each request is performed independently. In this context, the shipment separator unit (10) is divided into two categories, quantity-based or weightbased. Quantity-based shipments refer to pricing types where the entire shipment between locations from A to B is priced according to the customer's request. TL / Ton (currency / ton) shipments, on the other hand, are pricing types based on the weight equivalent to 1 ton between locations from A to B. After the choice between these two categories by the shipment separator unit (10), prediction is made within the scope of the invention method. Within the scope of the method, at least two, preferably three, different independent models are run. For this purpose, the selection of the model that specific to the requested pricing type is determined by the shipment separator unit (10).

[0049] The subject of the prediction method involves four artificial intelligence models, and the result obtained from these models is minimized with at least one mathematical model. Price prediction is carried out in conjunction with an algorithm operating on the decider unit (50).

[0050] The four artificial intelligence models operating in the artificial intelligence unit (20) which run in a continuous learning cycle. All models update themselves with newly arrived shipment results at predefined time intervals (e.g., every 24 hours). The data used by the artificial intelligence models include following datas:

[0051] • Shipment starting day

[0052] • Shipment starting month

[0053] • Shipment starting year

[0054] • Which working day of week

[0055] • Which working day of month

[0056] • Fuel prices (Currency)

[0057] • Departure location

[0058] • Destination location

[0059] • Distance (km)

[0060] • Number of stops (break / intermediate delivery area)

[0061] • Tonnage

[0062] • Vehicle type • Insurance ratio

[0063] • Currency exchange rate

[0064] Within the scope of the method, parameters such as seasonality, inflation rate, fuel prices, currency fluctuations, etc., are directly incorporated to prevent an increase in the error rate.

[0065] Within the artificial intelligence unit (20) where the method operates, there are four independent artificial intelligence modules (21 ). These artificial intelligence modules (21 ) run in artificial intelligence unit (20) which using LightGBM, ExtraTrees, XGBoost, and Random Forest algorithms independently from each other. The results produced independently by each artificial intelligence module (21 ) creates different error rates for adjacent days. At least one combinationer unit (30) evaluates these mentioned error rates and determines the best result, and it cancels out the other results by using the results in the artificial intelligence module (21 ) where that result was obtained. This way, it identifies only the best-performing data from the four models. These processes are refreshed in each learning cycle.

[0066] The invention includes at least one calibrator unit (40). When a shipment is created, mentioned calibrator unit (40), calculates prices based on filtering historical data according to the characteristics of that shipment by higher weight to the prices from the nearest dates based on averages and standard deviations during the calculation. During mentioned calculation that based on these averages and standard deviations, a pruning algorithm is run in at least one pruning module (41 ) to separate out the anomalous shipments from the calculation that have occurred between locations from A to B.

[0067] The pruning module (41 ) prevents the addition of pricing parameters to the general prediction in cases of a shipment other than the price for the route. (Except for the price associated with that route.) This module ensures that shipments that fall outside of the average +2 standard deviations of the general route are pruned, and the calibrator unit (40) is disabled for these shipments.

[0068] Figure 3 provides a flowchart that describing the operation of the pruning module (41 ). As seen in the diagram, in scenarios where the number of shipments is less than 3, the filters are gradually refreshed. This ensures that missing data (indicated in bold) is not considered. (In the case of rare shipments occurring for a specific shipment, the missing data is removed from the dataset.)

[0069] The invention includes a decider unit (50) that determines how much the data from the artificial intelligence unit (20) should be corrected with the data from the calibrator unit (40). The decider unit (50) operates within the said computer-aided system (S) works with bellow equation:

[0070] LenThe number of data points previously observed on this route.

[0071] KalThe result of the calibrator algorithm.

[0072] KomThe result of the combinationer algorithm. a : A constant

[0073] K: Decider

[0074] Thanks to the invention, the final freight price is estimated by dynamically with combining the artificial intelligence algorithm results from the combinationer unit (30) and the mathematical model results from the calibrator unit (40). Higher number of data points previously observed on this route is increases confidence in the estimation. Additionally, gradual corrections can be made based on the difference between the results from the combinationer unit (30) and the calibrator unit (40). This allows for the minimization of errors and deviations with taking sudden changes into account, and it estimates more accurate price which closer to reality.

[0075] The protection scope of the invention is defined in the attached claims and is not limited to the embodiments described in this detailed description. It is clear that a person skilled in the art can develop similar embodiments within the scope of the above-mentioned descriptions without deviating from the main theme of the invention.

Claims

CLAIMS1 . A method which operates on a computer-aided system (S) containing at least one processor and at least one memory unit, is designed to enable rapid and dynamic pricing between freight carriers and freight owners in road transportation characterized by;Separated the shipments based on the pricing per quantity or per ton: a. The data related to the shipment where the transportation process is carried out is predicted independently from each other by at least two artificial intelligence algorithms under the control of the mentioned processor; b. Determining the artificial intelligence algorithm with the lowest error rate among the results of the mentioned artificial intelligence algorithms under the control of at least one processor and generating results using the outcome of the best-performing algorithm; c. Taking the average with considering standard deviation; obtaining correction data by filtering the prices from past shipments based on the characteristics of that shipment, with a higher weight given to prices from the nearest dates shipments at the time of creating the shipment; d. Obtaining the related prediction result by multiplying the data results from the previously selected shipment with the result of the selected artificial intelligence algorithm and adding it to the result of the correction data algorithm.

2. The method according to the claim 1 , characterized by; pruning shipments that falling outside of the average +2 standard deviations of the general route with blocking the parameters specified in the c step to prevent their inclusion in the general prediction; in cases where a shipment has taken place outside of the pricing for the route,3. The method according to the claim 1 , characterized by; determination of the mentioned prediction result is in the d operation step with equation.

4. The method according to the claim 1 , characterized by; using at least two of the algorithms which from LightGBM, ExtraTrees, XGBoost, and Random Forest.

5. The method according to the claim 2, characterized by; pruning is carried out to ensure gradual renewal of filters, thereby preventing the consideration of missing data; in scenarios where the number of shipments is less than 3.

6. A computer-aided system (S) that containing at least one processor and at least one memory unit, is designed to enable rapid and dynamic pricing between freight carriers and freight owners in road transportation characterized in that; includes byA shipment separator unit (10) which separates the shipments based on the pricing per quantity or per ton;An artificial intelligence unit (20) containing at least two artificial intelligence modules (21 ) which enables independent prediction of data related to the shipment where the process is carried out;At least one combinationer unit (30) that determines the best-performing result from the mentioned two artificial intelligence modules (21 ) and transmits the outcome;A calibrator unit (40) that obtains correction data based on averages with taking into account standard deviations with higher weight to the prices from the nearest dates completed shipments by filtering historical shipment data according to the characteristics of that shipment which created at the time of the shipment;A pruning module (41 ) that is capable of disabling the mentioned calibrator unit (40) to prevent the inclusion of pricing parameters in the general prediction in the event that a shipment has taken place outside of the pricing related to the route;The decider unit (50) that determines the prediction result by reducing the error rate of the results from the calibrator unit (40) and the combinationer unit (30).

Images