JP7142838B2 - Shipping Market Prediction Operation System - Google Patents

Description

The present invention relates to a shipping market forecast operation system.

Marine freight rates in the shipping industry are greatly affected by global economic trends and the number of ships built. As the shipping industry fluctuates rapidly, it is important for shipping companies to forecast shipping market conditions and make management decisions based on the results of those forecasts in order to achieve sustainable profits.

Based on these characteristics, research on forecasting shipping market conditions has been conducted so far. For example, a method of forecasting shipping market conditions in the dry bulk market using a time-series analysis method has been proposed (Non-Patent Document 1). Moreover, in recent years, a method of forecasting shipping market conditions using a machine learning method has also been proposed (Non-Patent Document 2).

V. , Tsioumas, et al. : A Novel Approach to Forecasting the Bulk Freight Market. The Asian Journal of Shipping and Logistics, Vol. 33, Issue 1, 2017, pp. 33-41. Han, Q. , et al. : Forecasting dry bulk freight index with improved SVM. , Mathematical Problems in Engineering , 2014.

However, no research has been conducted to predict shipping market conditions by combining ship movement information and machine learning, and no technology has been established to predict shipping market conditions with high accuracy. does not exist.

According to claim 1 of the present invention, a shipping market forecast operation system for predicting and providing shipping market conditions includes ship data including information on past movement of the ship, shipping market data, and fuel used in the ship. Based on the fuel price data, the ship data, the shipping market data, and the fuel price data before the current time entered are applied to a prediction model constructed by machine learning using the shipping market data as training data. a shipping market prediction server having predictive analysis means for predicting shipping market conditions for a specified future date and/or specified period and outputting shipping market forecast results; and storage means for storing the shipping market forecast results; A shipping market forecast operation system characterized by comprising data providing means for enabling viewing of the shipping market forecast results via a communication line and/or a network.

Here, the predictive analysis means performs the machine learning based on the market price data of the cargo to be loaded on the ship as the shipping market data to construct the prediction model, and the input market price of the cargo. Preferably, data is applied to said forecast model to forecast said shipping market conditions.

Further, the predictive analysis means performs the machine learning based on the stock price index data of industries related to the cargo to be loaded on the ship as the shipping market data , constructs the prediction model, and performs the input stock price. Preferably, index data is applied to the forecast model to forecast the shipping market conditions.

In addition, it is preferable that the predictive analysis means change the prediction model for predicting the shipping market forecast results in the predictive analysis means based on information acquired from a user browsing the shipping market forecast results. be.

The ship data preferably includes data obtained from an automatic ship identification system (AIS) and individual ship data.

Further, it is preferable that the shipping market data is the Baltic Shipping Index and/or ship type data that is the basis for calculation of the Baltic Shipping Index.

Further, it is preferable that the shipping market forecast result is a forecast result for a specified future date and/or a specified period.

Further, it is preferable that the predictive analysis means and the data providing means possessed by the developer of the predictive analysis means provide viewing of the shipping market forecast results.

Further, it is preferable that the predictive analysis means and the data providing means possessed by a provider of data used by the predictive analysis means provide the browsing of the shipping market forecast results.

Moreover, it is preferable that the data provider is a provider of the ship data.

Further, it is preferable that the data provider is a provider of the shipping market data. Further, it is preferable that the data provider is a provider of the fuel price data. Further, it is preferable that the data provider is a provider of market price data of the freight. Moreover, it is preferable that the data provider is a provider of the stock index data.

Further, it is preferable that the forecast analysis means and the data provision means possessed by a business operator other than the data provider provide the browsing of the shipping market forecast result.

In addition, it is preferable that the predictive analysis means develops a construction method of the predictive model by a developer of the predictive analysis means, and constructs the predictive model by the data provider. In addition, it is preferable that the predictive analysis means has a predictive model construction method developed by a developer of the predictive analysis means, and the predictive model constructed by a business operator other than the data provider. be.

According to the shipping market forecast operation system for predicting and providing shipping market conditions for ships according to claim 1 of the present invention, ship data including information on past movement of the ship, shipping market data, and Based on the fuel price data of the fuel , the above-mentioned ship data, the above-mentioned shipping market data, and the above-mentioned fuel price data that have been input before the current time are input to a prediction model constructed by machine learning using the above-mentioned shipping market data as training data. A shipping market prediction server having predictive analysis means for predicting shipping market conditions for a specified future date and/or specified period and outputting shipping market forecast results, and storage means for storing the shipping market forecast results. and a data providing means that enables viewing of the shipping market forecast results via a communication line and/or network, thereby providing users with indicators of shipping market conditions predicted with high accuracy. .

Further, the predictive analysis means constructs the predictive model by performing the machine learning based on the market price data of the cargo to be loaded on the ship as the shipping market data , and inputs By applying the market price data of the cargo to the prediction model to predict the shipping market conditions, the shipping market conditions can be predicted with higher accuracy.

Further, the predictive analysis means performs the machine learning based on the stock price index data of industries related to the cargo to be loaded on the ship as the shipping market data , constructs the prediction model, and performs the input stock price. By applying the index data to the prediction model to predict the shipping market conditions, the shipping market conditions can be predicted with higher accuracy.

Further, the predictive analysis means changes the prediction model for predicting the shipping market forecast results in the predictive analysis means based on the information acquired from the user browsing the shipping market forecast results. It is possible to build a forecast model that meets the needs of the market, and forecast the shipping market with a higher degree of accuracy.

Further, the ship data includes data acquired from an automatic ship identification system (AIS) and individual ship data, so that shipping market conditions can be predicted with higher accuracy.

Further, the shipping market data is the Baltic Shipping Index and/or ship-type data on which the Baltic Shipping Index is calculated. It can be predicted with high accuracy.

Further, since the shipping market forecast result is a forecast result for a future specified date and/or specified period, it is possible to forecast the shipping market condition for a specified future date and/or specified period with high accuracy.

In addition, the forecast analysis means and the data provision means possessed by the developer of the forecast analysis means provide the result of the forecast of shipping market conditions, so that the forecast result of the shipping market is provided by the developer of the forecast analysis means. A service for viewing the results can be provided.

In addition, the forecast analysis means and the data providing means possessed by the provider of the data used in the forecast analysis means provide the result of forecasting shipping market conditions, thereby providing the data to be used in the forecast analysis means. can provide a service for browsing the shipping market forecast results from the provider of the above.

Further, the data provider, being the ship data provider, can provide a service that allows the ship data provider to browse the shipping market forecast results.

In addition, the data provider is the shipping market data provider, or the data provider is the fuel price data provider, or the data provider is , being a provider of the cargo market price data, or being a provider of the stock price index data , the data provider receives the shipping market forecast results from each data provider. It is possible to provide browsing services.

Further, by providing viewing of the shipping market forecast results by the predictive analysis means and the data providing means possessed by a business operator other than the data provider, It is possible to provide a service for viewing market forecast results.

In addition, the predictive analysis means has a predictive model construction method developed by a developer of the predictive analysis means, and the predictive model is constructed by the data provider. It is possible to provide a service that allows users to browse the shipping market forecast results forecasted by their own forecasting model. In addition, in the predictive analysis means, the predictive model construction method is developed by a developer of the predictive analysis means, and the predictive model is constructed by a business operator other than the data provider. It is possible to provide a service that allows users to browse the shipping market forecast results forecasted by a forecasting model constructed by a company other than the data provider itself.

It is a figure which shows the structure of the shipping market forecast in embodiment of this invention. BRIEF DESCRIPTION OF THE DRAWINGS It is a functional block diagram which shows the structure of the shipping market forecast operation system in embodiment of this invention. It is a figure which shows the structure of the shipping market forecast server and data provision means in embodiment of this invention. It is a flowchart which shows the shipping market forecast method in embodiment of this invention. It is a figure which shows the conditions example of the sea area which the ship in embodiment of this invention is navigating. FIG. 4 is a diagram for explaining data selection processing according to the embodiment of the present invention; It is a figure which shows the model example of the deep neural network (DNN) in embodiment of this invention. It is a figure which shows the model example of the deep neural network (DNN) in embodiment of this invention. It is a figure for demonstrating the forecast result of the index of the shipping market condition in embodiment of this invention. It is a figure which shows the example of conditions of prediction model construction in embodiment of this invention. It is a figure which shows the example of a time-series prediction result in embodiment of this invention. FIG. 5 is a diagram showing the correlation and direction matching rate between actual values and predicted values for prediction results in the embodiment of the present invention; FIG. 4 is a diagram for examining the influence of input data on prediction in the embodiment of the present invention; FIG. BRIEF DESCRIPTION OF THE DRAWINGS It is a figure which shows the structure of the shipping market forecast operation system in embodiment of this invention. BRIEF DESCRIPTION OF THE DRAWINGS It is a figure which shows the structure of the shipping market forecast operation system in embodiment of this invention. BRIEF DESCRIPTION OF THE DRAWINGS It is a figure which shows the structure of the shipping market forecast operation system in embodiment of this invention. It is a flowchart which shows the data acquisition method in the shipping market forecast operation system in embodiment of this invention. It is a flowchart which shows the data acquisition method in the shipping market forecast operation system in embodiment of this invention. 4 is a flow chart showing a setting method for forecasting shipping market conditions in the shipping market conditions forecasting operation system according to the embodiment of the present invention.

A shipping market forecasting system 100 according to the embodiment of the present invention comprises a shipping market forecasting server 102, data providing means 104, and a client computer 106, as shown in FIG.

Shipping market prediction server 102 and client computer 106 are connected via data providing means 104 and network 108 so that information can be exchanged. The network 108 may be wired or wireless, and is, for example, a public line such as the Internet.

As shown in FIG. 2, the shipping market forecast server 102 includes data acquisition means 10, cleansing means 12, data complementation means 14, data extraction means 16, data leveling means 18, data selection means 20, predictive analysis means 22, time It comprises sequence processing means 24 and display means 26 .

The shipping market prediction server 102 and the data providing means 104 can be implemented by a computer including a processing unit 30, a storage unit 32, an input unit 34, an output unit 36 and a communication unit 38, as shown in FIG. can.

The processing unit 30 includes a CPU and the like, and performs processing in the shipping market forecast system 100 in an integrated manner. The processing unit 30 executes the shipping market forecasting program stored in the storage unit 32, thereby realizing the shipping market forecasting method according to the present embodiment. That is, the computer functions as data acquisition means 10, cleansing means 12, data complementation means 14, data extraction means 16, data leveling means 18, data selection means 20, predictive analysis means 22, time-series processing means 24, and display means 26. Let The storage unit 32 stores information such as a shipping market prediction program used in shipping market prediction processing in the shipping market prediction system 100 and data used for learning. The storage unit 32 can be composed of, for example, a semiconductor memory, a hard disk, or the like. The input unit 34 includes means for inputting information to the shipping market forecast system 100 . The input unit 34 includes, for example, a keyboard, a mouse, and the like. The output unit 36 includes means for displaying information processed by the shipping market forecast system 100 . The output unit 36 can be, for example, a display. The communication unit 38 includes an interface for exchanging information with an external device. The communication unit 38 enables communication with an external device by being connected to, for example, an information communication network such as the Internet or a dedicated line.

Hereinafter, the shipping market forecasting method in the shipping market forecasting system 100 will be described along the flowchart of FIG. The shipping market forecast system 100 of the present embodiment uses ship data including information on the movements of ships, shipping market data, and fuel price data of fuel used in ships to generate an index indicating shipping market conditions related to ships. Do something to predict.

In step S10, data acquisition processing is performed. By the processing of this step, the shipping market forecast system 100 functions as the data acquisition means 10 . The processing unit 30 acquires data used in shipping market forecast processing. The data may be input by the user via the input unit 34 or may be obtained from the network 108 from an external device or the like via the communication unit 38 .

Data for constructing a prediction model of indices related to shipping market conditions in the shipping market forecasting system 100 includes ship data including information on movement of ships, shipping market data, and fuel price data for fuel used in ships. The data shall be time-series data showing changes over time.

Vessel data, including information about vessel movement, may be obtained from Automatic Identification System (AIS) information. AIS is a kind of broadcast-type automatic slave surveillance (ADS-B), and it regularly transmits voyage information such as ship name and ship position by radio for the purpose of improving the safety of ship navigation. It is designed to receive voyage information such as the name of the ship and the position of the ship from the ship by radio. AIS information includes static information, dynamic information, navigation-related information, and navigation safety-related information. Static information includes information necessary to identify a vessel, such as vessel name and call sign. The dynamic information includes information on the ever-changing voyage situation of the vessel (information such as current position, voyage speed, course, etc.). Specifically, AIS information includes ship name, MMSI number, IMO number, ship type, ship length/width, static information such as GNSS antenna position, ship position, UTC, course over ground (COG), Navigation-related information such as speed over ground (SOG), heading direction, turning angular velocity, dynamic information on voyage conditions (during voyage, anchoring, berthing, etc.), draft, cargo, destination, estimated time of arrival, etc. include. AIS information can be obtained as time-series data. The transmission interval of the AIS information varies depending on the speed of the ship, but it is acquired at intervals of 3 minutes at maximum and at least 2 seconds at minimum.

Vessel data in the present embodiment may include information selected from static information and dynamic information, or information obtained based on such information. For example, it is preferable to use the ship's course over ground (COG) and ship's speed over ground (SOG) for each time as ship data.

Further, in the data acquisition process, individual ship data is acquired. The individual vessel data preferably includes the vessel name, vessel ID, and vessel deadweight tonnage (DWT). Individual ship data can be obtained from Astra-paging individual ship data. Based on the individual ship data, the deadweight tonnage (DWT) of the ship corresponding to the ship name included in the ship data can be obtained.

Further, in the data acquisition process, shipping market data is acquired. Shipping market data is information on shipping market conditions, such as freight rates for ships in shipping. Shipping market data is used as teacher data for machine learning to predict future values of indices indicating shipping market conditions in the shipping market forecast system 100 .

Shipping market data includes, for example, the Baltic Shipping Index, which is a freight rate index for ocean-going tramp vessels announced by the Baltic Shipping Exchange in London. The Baltic Shipping Index is provided each day the exchange is open. In addition, the Baltic Shipping Index is subdivided into vessel type data based on which calculations are made, such as comprehensive type (BDI), Cape type (BCI), Panamax type (BPI), Supramax type (BSI), Handysize type (BHSI), etc. be done. Therefore, when using the Baltic shipping index as shipping market data, one of these subdivided indices should be selected.

Note that the shipping market data is not limited to the Baltic shipping index, and may be related to the shipping market conditions (indices) to be predicted by the shipping market prediction system 100, such as the tanker freight index, the container freight index, etc. Anything is fine.

As a specific example, when predicting an index that indicates the shipping market conditions of ships that carry iron ore, the cape type (BCI) of the Baltic shipping index is obtained.

In addition, in the data acquisition process, information on the price and market of cargo to be loaded in the shipping market conditions to be predicted is acquired. For example, when predicting an index indicating the maritime market conditions of ships carrying iron ore, the price of iron ore is obtained as the price of the cargo, and stock price index data relating to the steel industry is obtained as the information on the market of the cargo. Also, for example, when predicting an index indicating the shipping market conditions of ships carrying grain, the price of the grain is obtained as the price of the cargo, and the stock price index data of the grain industry is obtained as the cargo market information. This information is provided each day the exchange is open.

As a specific example, when predicting an index that indicates the shipping market conditions for ships carrying iron ore, daily data on iron ore fines and 62% Fe CFR futures prices are obtained as iron ore prices, and the iron ore market Tokyo SE TOPIX17 Steel Stock Price daily data is acquired as information related to

Also, in the data acquisition process, fuel price data of the fuel used in the ship is acquired. If the fuel used in the ship is crude oil, for example, a WTI price is obtained that indicates the price of crude oil. If the vessel uses a fuel other than crude oil, information is obtained indicating the price of that fuel. This information is provided each day the exchange is open.

Data on the expected date (designated date) and period may be input by the user, for example, from the data input means 10a provided in the data acquisition means 10. FIG.

In step S12, cleansing processing is performed on the data. By the processing of this step, the shipping market forecast system 100 functions as the cleansing means 12 . The processing unit 30 performs cleansing processing such as removing unnecessary data from the data acquired in step S10 and adjusting the density of the data.

AIS information may contain noise in the data due to effects such as different ships using the same MMSI number. In addition, AIS information is time-series data at intervals of several seconds to several minutes, and the amount of information may become excessive. Therefore, a process of thinning out data time intervals is performed. As a specific example, the process of calculating the average value of data values every 10 minutes so that the time interval of the course over ground (COG) and speed over ground (SOG) of the ship obtained from AIS information is 10 minutes. conduct. Also, the data with the ground speed (SOG) of 30 knots or more is treated as noise and removed.

Note that the data cleansing process is not limited to these processes, and may be any process that adjusts the data in accordance with the predictive analysis process performed by the predictive analysis means 22, which will be described later.

In step S14, a complementary process is performed on the data. By the processing of this step, the shipping market forecasting system 100 functions as the data complementing means 14 . The processing unit 30 performs a process of complementing the data acquired in step S10 when necessary data is missing.

For example, with regard to individual ship data, when data is missing, it is supplemented with data provided by the ITU, etc. In addition, processing is performed to supplement the data on days when the stock exchange is closed, such as shipping market data, information on the prices and markets of cargo to be loaded, and fuel price data on fuel used in ships. As a specific example, the data of the previous day is used as the data of the day when the exchange is closed.

Note that the data complementing process is not limited to these processes, and may be any process that complements the data in accordance with the predictive analysis process performed by the predictive analysis means 22, which will be described later.

In step S16, a process of extracting data is performed. By the processing of this step, the shipping market forecast system 100 functions as the data extracting means 16 . The processing unit 30 performs a process of extracting necessary data from the data subjected to the data cleansing process and the data complementing process.

The extraction condition is a condition for extracting data that can appropriately construct a prediction model by machine learning in the prediction analysis process described later. Specifically, for example, the type of vessel, the sea area and period in which the vessel is navigating, and conditions related to the movement of the vessel. The condition of the type of vessel can be a condition related to dead weight tons (DWT), size of the vessel (ship length, width, etc.), and the like. The condition of the sea area in which the ship is navigating can be a sea area related to the predicted shipping market condition index. The period during which the ship is sailing can be a period corresponding to the period for which it is desired to predict the index of the shipping market conditions. Also, the conditions relating to the movement of the vessel can be the range of speed over ground (SOG) and the range of course over ground (COG) of the vessel.

When forecasting indicators that show the shipping market conditions for ships carrying iron ore, as shown in Figure 5, the conditions for the sea areas where ships are navigating are the coastal waters of Australia and the Indian Ocean region through which iron ore-carrying ships pass. Extract data only for vessels transiting the southern waters. Also, only data relating to vessels carrying iron ore with a deadweight tonnage (DWT) of 180,000 DWT or more is extracted. In addition, as a condition related to the motion of the ship, it is preferable that the course over ground (COG) of the ship is extracted from the direction and speed range of the ship loaded with iron ore in each sea area. Specifically, the ground speed (SOG) is 3 knots or more, and the course over the ground (COG) is 80 degrees or less and 270 degrees or more in the waters near Australia, and the course over the ground (COG) is 10 degrees or more in the waters south of India. Only extract data for vessels below 170°. Also, for example, only data from August 2014 to January 2017 is extracted as a period condition according to the expected date.

In addition, when the unit of the period of each data is different, processing is performed to match the unit of those periods. For example, if shipping market data and fuel price data are daily data, ship data is adjusted to daily data accordingly. Specifically, the ship data extracted based on the above extraction conditions is subjected to an averaging process to obtain daily data. For example, the ship speed of each ship is averaged to calculate the daily average ship speed. Also, the deadweight tonnage (DWT) of the vessels that meet the extraction conditions are summed up to calculate the total deadweight tonnage (total DWT) indicating the total amount of cargo for each day.

In step S18, a process of leveling the data is performed. By the processing of this step, the shipping market forecast system 100 functions as the data leveling means 18 . The processing unit 30 performs a process of leveling the data by temporal data averaging process or the like. For example, since the data obtained by the above processing includes daily fluctuations, a 14-day moving average is calculated for each day to level the data.

In step S20, a process of selecting data is performed. By the processing of this step, the shipping market forecast system 100 functions as the data selection means 20 . The data extracted in step S16 and leveled in step S18 may include a period in which data is missing, as indicated by the dashed circle in FIG. For example, gross deadweight tonnage (gross DWT) and vessel ground speed (SOG) data obtained from AIS information may lack data for periods when the satellite is stationary. A process is then performed to remove other data for those periods. In the example of FIG. 6, processing is performed to leave only the data of the period P1 and the period P2 and remove the data of the other periods.

The data complementing process in step S16, the data leveling process in step S18, and the data selection process in step S20 may be performed as required, and are not essential processes.

In step S22, a predictive analysis process is performed to predict an index indicating shipping market conditions related to the ship. By the processing of this step, the shipping market forecast system 100 functions as the forecast analysis means 22 . The processing unit 30 uses the data prepared up to step S20 to construct a prediction model for predicting an index indicating the shipping market condition related to the ship by machine learning.

Deep learning (DL) can be applied as machine learning. Deep learning (DL) includes techniques such as deep neural network (DNN), recurrent neural network (RNN), support vector machine (SVM), and XG-Boost. In the present embodiment, the model to be applied is not particularly limited, and may be appropriately selected according to the index of the shipping market conditions to be predicted and the input data (feature vector).

As a specific example, a deep neural network (DNN) was used to build a model that predicts the future Cape type (BCI) of the Baltic shipping index. As shown in FIG. 7, a deep neural network (DNN), which is simply a multi-layered neural network, was used. Linear was used as the activation function of the input layer and the output layer, and ReLU was applied to the activation function of the intermediate layer. Also, at the time of learning, Dropout was incorporated for all hidden layers. The dropout rate (Dropout rate) was set to 0.2. We also incorporated Batch Normalization in all layers. For the learning rate and the like, the values in the reference (Kingma, D., & Ba, J.: Adam: A method for stochastic optimization, 2014, arXiv preprint arXiv: 1412.6980.) were used.

ここで、ΔＢＣＩ：ＢＣＩの増減分（ｉｎｄｅｘ），ｆ：ＢＣＩの増減予測モデル，Ｄ：入力データの集合（詳細はＴａｂｌｅ １を参照），ΔＤ：入力データの差分値，Ｓ：統計データの集合，Ｉｏ：取得した衛星ＡＩＳデータ（インド洋），Ａｕ：取得した衛星ＡＩＳデータ（オーストラリア），ｔ：現在の日（ｄａｙ），ｐｔ：ｐｔ日先の日（ｄａｙ）（本具体例では、３日，７日，１４日，３０日） Expressions (1) to (3) show the relationship of the input data in the prediction model.

Here, ΔBCI: BCI increase/decrease (index), f: BCI increase/decrease prediction model, D: set of input data (see Table 1 for details), ΔD: difference value of input data, S: set of statistical data , Io: acquired satellite AIS data (Indian Ocean), Au: acquired satellite AIS data (Australia), t: current day (day), pt: pt days ahead (day) (in this specific example, 3 day, 7th, 14th, 30th)

A data group (feature vector) shown in FIG. 8 was used as specific input data. That is, the input data D is a data group (feature vector) in which the data shown in FIG. 8 are combined for each day. Data for the past 60 days (t-60) from the present time (t) were used as input data. Furthermore, the difference value ΔD from the previous day (how much the current value increased or decreased compared to the previous day's value) was also added to the input data. The difference value is a difference value for each type of data.

However, the input data is not limited to the data group (feature vector) shown in FIG. Anything that contains Furthermore, the input data may include information on the price of cargo (for example, iron ore, etc.) in the shipping market conditions to be forecasted, and market information related to the shipping market conditions to be forecasted.

Output data, as shown in FIG. 9, was the increase/decrease of the future Baltic Shipping Index (BCI) for 3 days, 1 week, 2 weeks, and 1 month ahead. The setting/change of the predicted date is performed by the variable section 22a provided in the prediction analysis means 22. FIG.

Specifically, under the conditions shown in FIG. 10, machine learning was performed by applying supervised data in a deep neural network (DNN). However, the activation functions and learning conditions of the input layer, intermediate layer, and output layer are not limited to the above examples, and are preferably optimized as appropriate.

In step S24, time-series processing is performed on the output data. By the processing of this step, the shipping market forecast system 100 functions as the time-series processing means 24 . Using the above prediction model, predictive analysis of indicators of shipping market conditions is performed while changing forecast dates to generate time-series data of predicted values of indicators of shipping market conditions.

In a specific example, as shown in FIG. 11, while changing the forecast date, forecast values for three days, one week, two weeks, and one month ahead were obtained to obtain time-series data. In FIG. 11, the solid line is the result of building a prediction model by machine learning including ship data related to the movement of ships obtained from AIS information in the input data (feature vector) (Case 1), and the dashed line is the input data (feature vector ) shows the results when a prediction model is constructed by machine learning without including ship data on the movement of ships obtained from AIS information (Case 2). In addition, in FIG. 11, the dotted line indicates the actual value.

In the case of Case 1, the temporal trend of the Cape type (BCI) of the Baltic Shipping Index was successfully reproduced one month and two weeks ahead. On the other hand, in Case 2, the temporal trend of the Cape type (BCI) of the Baltic Shipping Index could not be sufficiently reproduced in any case.

FIG. 12 shows the result of examining the correlation between the actual value and the predicted value and the directional matching rate. The directional matching rate is an index that indicates the matching rate of the direction of movement between the prediction result and the actual value. The prediction results for one month ahead, two weeks ahead, and one week ahead show a directional coincidence rate of over 70%. This was a higher value than Case2. In Case 1, the correlation coefficient was 0.5 or more for one month ahead, two weeks ahead, and one week ahead, confirming that the correlation with the actual value is high.

In the above specific example, the effect of input data on the prediction result of the Cape type (BCI) of the Baltic Shipping Index was examined. FIG. 13 shows the prediction results when all of the input data shown in FIG. The forecast results are shown when the stone price is not included in the input data. The prediction results show the directional coincidence rates and correlations in predictions one month ahead and two weeks ahead.

When the stock price index was not included in the input data, both the correlation coefficient between the actual and predicted values and the directional agreement decreased. However, the directional matching rate was only slightly lowered. Moreover, when the crude oil price was not included in the input data, both the directional agreement rate and the correlation coefficient between the actual and predicted values decreased significantly. When the iron ore price was not included in the input data, the directional agreement rate and correlation coefficient between the actual values and the predicted values decreased slightly in the one-month ahead forecast, but the directional agreement in the two-week ahead forecast The concordance rate and correlation coefficient were greatly improved.

Based on these results, information on iron ore prices (price of cargo to be loaded) may be excluded from the input data. However, it may be better to include it in the input data if the predicted shipping market conditions or the cargo to be loaded are changed. Stock indices may also be better included in the input data to improve correlations.

As described above, according to the present embodiment, by performing machine learning using input data (feature vectors) including ship data including information on ship movements, it is possible to predict indicators related to shipping market conditions with high accuracy. . In particular, by performing machine learning using input data (feature vectors) that combine ship data including information on ship movement, shipping market data, and fuel price data for fuel used in ships, we can develop indicators related to shipping market conditions. can be predicted with higher accuracy. If we focus only on accuracy, if we consider the stock price index (stock price index data) and iron ore prices (freight market price data), the accuracy will increase, and it is possible that more suitable data for forecasting will emerge in the future. There is also sex. However, considering the acquisition price of data and the time required for prediction processing, etc., it is preferable to narrow down the input data to data that is highly likely to affect accuracy.

In step S26, output processing of the prediction result is performed. By the processing of this step, the shipping market forecast system 100 functions as the display means 26 . The processing unit 30 outputs the prediction result (time-series data, etc.) obtained by the above processing through the output unit 36 .

The forecast result can also be output to the outside of the shipping market forecast server 102 using the data providing means 104 . The output means 40 of the data providing means 104 is connected to the network 108 and transmits the time-series prediction results generated by the time-series processing means 24 to the client computer 106 . Also, the communication means 42 of the data providing means 104 transmits the time-series prediction results generated by the time-series processing means 24 to the outside through a dedicated line or the like.

[Marine transport market forecast operation system]
Hereinafter, a shipping market forecast operation system 200 that provides users with prediction results of shipping market conditions using the shipping market forecast system 100 described in the above embodiment will be described.

FIG. 14 shows the configuration of the shipping market forecast operation system 200. As shown in FIG. In this configuration, a company other than the data provider operates the shipping market prediction server 102 and the data providing means 104 . In this example, the forecast analysis means developer who developed the forecast analysis means 22 of the shipping market forecast server 102 operates the shipping market forecast server 102 and the data providing means 104 . The shipping market forecast server 102 receives data from ship data providers, shipping market data providers, fuel price data providers, cargo market price data providers, and stock index data providers that are not predictive analysis means developers. Obtain ship data, shipping market data, fuel price data, cargo market price data and price index data, respectively. Ship data providers, shipping market data providers, fuel price data providers, cargo market price data providers, and stock index data providers should each have a server for providing data. good. Then, the predictive analysis means 22 of the shipping market prediction server 102 uses these data to predict the indices of the shipping market, and the data providing means 104 uses the shipping market prediction results from customers (customers A to C) who are users. Make it viewable. When the predictive analysis means developer operates the shipping market forecast server 102 and the data providing means 104, the shipping market forecast server 102 is changed to obtain forecast results with changed input data, or the variable part 22a is changed. It is possible to respond quickly.

FIG. 15 shows another configuration of the shipping market forecast operation system 200. As shown in FIG. In this configuration, the data provider operates the shipping market prediction server 102 and the data providing means 104 . This example shows an example in which a ship data provider operates the shipping market prediction server 102 and the data providing means 104 . The shipping market forecast server 102 receives provision of the predictive analysis means 22 used in the shipping market forecast server 102 from a predictive analysis means developer who is not a ship data provider. In addition, the shipping market forecast server 102 receives shipping market data, fuel price data, and cargo market price from shipping market data providers, fuel price data providers, cargo market price data providers, and stock index data providers, respectively. Get data and price index data. Ship data is provided by ship data providers themselves. At this time, the shipping market data provider, the fuel price data provider, the cargo market price data provider, and the stock index data provider should each have a server for providing data. Then, the predictive analysis means 22 of the shipping market prediction server 102 uses these data to predict the indices of the shipping market, and the data providing means 104 uses the shipping market prediction results from customers (customers A to C) who are users. Make it viewable. When a ship data provider operates the shipping market forecast server 102 and the data providing means 104, it can effectively utilize the wide variety of ship data acquired from a global perspective to obtain forecast results, Shipping market forecast results can be delivered together with ship data to customers who provide information.

FIG. 16 shows another configuration of the shipping market forecast operation system 200. As shown in FIG. In this example, a shipping market data provider operates the shipping market prediction server 102 and the data providing means 104 . The shipping market forecast server 102 is provided with the predictive analysis means 22 used in the shipping market forecast server 102 from a predictive analysis means developer who is not a shipping market data provider. In addition, the shipping market forecast server 102 receives ship data, fuel price data, cargo market price data and Get price index data. Shipping market data is provided by shipping market data providers themselves. At this time, the ship data provider, the fuel price data provider, the freight market price data provider, and the stock price index data provider should each have a server for providing data. Then, the predictive analysis means 22 of the shipping market prediction server 102 uses these data to predict the indices of the shipping market, and the data providing means 104 uses the shipping market prediction results from customers (customers A to C) who are users. Make it viewable. When the shipping market data provider operates the shipping market prediction server 102 and the data providing means 104, in addition to providing the current shipping index, it is possible to distribute the future prediction results of its own shipping index as a set. can.

Similarly, instead of the ship data provider, the shipping market data provider, the fuel price data provider, the cargo market price data provider, and the stock price index data provider provide the shipping market forecast server 102 and the data providing means 104. may be configured to operate.

At this time, the predictive analysis means developer who developed the predictive analysis means 22 of the shipping market forecast server 102 develops a predictive model construction method, gives the predictive model construct method to the data provider, and the predictive model is It may be constructed by a data provider. In addition, the predictive analysis means developer who developed the predictive analysis means 22 of the shipping market prediction server 102 develops a predictive model construction method, and provides the predictive model construction method to a business operator other than the data provider, The prediction model may be constructed by the business operator concerned.

Here, as shown in FIG. 17, in order to acquire data from a ship data provider, a shipping market data provider, a fuel price data provider, a freight market price data provider, and a stock index data provider, may be given to the shipping market prediction server 102, and various data may be transmitted to the shipping market prediction server 102 when authentication is performed using the access key.

Specifically, when the shipping market forecast server 102 requests the data provider server that manages the access key to use the data acquisition service (step S30), the access key is issued to the data provider server. A selection is made (step S32), and an access key is issued to the shipping market forecast server 102 (step S34). The shipping market forecast server 102 may perform processing such as displaying the issued access key (step S36). Next, when the access key is entered (step S38), data acquisition is set in the shipping market forecast server 102 (step S40), and the access key is transmitted to the data provider server (step S42). . In the data provider server, the access key is received, and if the access key is correct, authentication for providing data is established (step S44). After that, when the shipping market prediction server 102 requests data transmission (step S46), the request is processed in the data provider server (step S48), and the data is sent to the shipping market prediction server 102 (step S48). step S50). When the shipping market forecast server 102 receives the data through such processing, it performs a shipping market forecasting process according to the flowchart shown in FIG.

Further, as shown in FIG. 18, the Internet may be accessed from the shipping market forecast server 102, and various data may be acquired by scraping using a data acquisition program (scraping program).

Specifically, when data acquisition is set in the shipping market forecast server 102 (step S60), a data acquisition program (scraping program) is set (step S62) and started (step S64). As a result, a public communication network such as the Internet is accessed, and data is extracted and the extracted data is obtained (step S66). When the shipping market forecast server 102 receives the data through such processing, it performs a shipping market forecasting process according to the flowchart shown in FIG.

Also, as shown in FIG. 19, an access key is assigned to a user who intends to receive the shipping market forecast results from the shipping market forecast server 102, and authentication is performed using the access key entered at the user terminal. In this case, the shipping market prediction result may be provided from the shipping market prediction server 102 to the user terminal.

Specifically, when the user terminal requests the shipping market prediction server 102, which manages access keys, to use the service (step S70), the shipping market prediction server 102 selects an access key. (Step S72), an access key is issued to the user terminal (Step S74). The user terminal may perform processing such as displaying the issued access key (step S76). Next, when the access key is entered, the access key is transmitted from the user terminal to the shipping market prediction server 102 (step S78). The shipping market prediction server 102 receives the access key, and if the access key is correct, authentication for predicting the shipping market index is established (step S80). After that, if there is a request from the user terminal to change the setting of the predictive analysis (step S84), the setting of the predictive analysis means 22 in the shipping market prediction server 102 is changed. For example, when a request is made to set a date and a period for forecasting an index of shipping market conditions, the forecast date and period are set by the variable part 22a provided in the forecast analysis means 22. FIG. Further, for example, when the setting of the conditions of the prediction model shown in FIG.

As described above, according to the shipping market forecast operation system 200, it is possible to provide users with the prediction results of the shipping market forecasted by the shipping market forecast system 100. FIG.

In the above embodiment, the prediction of the Cape type (BCI) predictive analysis process of the Baltic Shipping Index (BCI) for iron ore and the provision of the prediction result were described, but the scope of application of the present invention is not limited to this. It can also be applied to predictive analysis of other indicators of market conditions and to provide predictive results.

10 data acquisition means, 10a data input means, 12 cleansing means, 14 data complementation means, 16 data extraction means, 18 data leveling means, 20 data selection means, 22 prediction analysis means, 22a variable section, 24 time series processing means, 26 display means, 30 processing unit, 32 storage unit, 34 input unit, 36 output unit, 38 communication unit, 40 output means, 42 communication means, 100 shipping market forecast operation system, 102 shipping market forecast server, 104 data providing means, 106 client computer, 108 network, 200 shipping market prediction operation system.

Claims (16)

A marine transportation market prediction operation system for predicting and providing shipping market conditions for ships,
A prediction model constructed by machine learning using the shipping market data as training data, based on ship data including information on past movement of the ship, shipping market data, and fuel price data of the fuel used in the ship. , apply the input ship data, shipping market data, and fuel price data before the current time, predict shipping market conditions for a specified future date and/or specified period, and obtain shipping market forecast results. predictive analytics means for output;
a shipping market prediction server having storage means for storing the shipping market prediction results;
A shipping market forecast operation system, comprising: data providing means for enabling viewing of the shipping market forecast results via a communication line and/or a network. The shipping market forecast operation system according to claim 1,
The predictive analysis means builds the predictive model by performing the machine learning based on the market price data of the cargo to be loaded on the ship as the shipping market data , and uses the input market price data of the cargo as the A shipping market forecast and operation system, characterized by applying a prediction model to predict the shipping market conditions. The shipping market forecast operation system according to claim 1 ,
The predictive analysis means constructs the predictive model by performing the machine learning based on the stock price index data of the industry related to the cargo to be loaded on the ship as the shipping market data , and inputs the stock price index data. to the prediction model to predict the shipping market conditions. The shipping market forecast operation system according to any one of claims 1 to 3 ,
Said predictive analysis means changes said prediction model for predicting said shipping market forecast results in said predictive analysis means, based on information obtained from a user viewing said shipping market forecast results. Market forecast operation system. The shipping market forecast operation system according to any one of claims 1 to 4 ,
A shipping market prediction operation system, wherein the ship data includes data acquired from an automatic ship identification system (AIS) and individual ship data. The shipping market forecast operation system according to any one of claims 1 to 5 ,
A shipping market forecast operation system, wherein the shipping market data is a Baltic shipping index and/or ship type data that is a basis for calculation of the Baltic shipping index. The shipping market forecast operation system according to any one of claims 1 to 6 ,
A shipping market forecast operation system, wherein the forecast analysis means and the data providing means possessed by a developer of the forecast analysis means provide access to the forecast result of the shipping market. The shipping market forecast operation system according to any one of claims 1 to 6 ,
A shipping market forecast operation system, wherein the forecast analysis means and the data providing means possessed by a provider of data used by the forecast analysis means provide viewing of the shipping market forecast results. The shipping market forecast operation system according to claim 8 ,
A shipping market forecast operation system , wherein the data provider is a provider of the vessel data. The shipping market forecast operation system according to claim 8 ,
A shipping market forecast operation system, wherein the data provider is a provider of the shipping market data. The shipping market forecast operation system according to claim 8 ,
The shipping market forecast operation system, wherein the data provider is a provider of the fuel price data. The shipping market forecast operation system according to claim 8 , which is dependent on claim 2,
The shipping market forecast operation system, wherein the data provider is a provider of market price data for the cargo. The shipping market forecast operation system according to claim 8 , which is dependent on claim 3 ,
The shipping market forecast operation system, wherein the data provider is a provider of the stock index data. The shipping market forecast operation system according to any one of claims 8 to 13 ,
A shipping market forecast and operation system, wherein the forecast analysis means and the data providing means owned by a business operator other than the data providing business operator provide browsing of the shipping market forecast results. The shipping market forecast operation system according to any one of claims 8 to 13 ,
A shipping market forecast operation system, wherein the predictive analysis means is constructed by a developer of the predictive analysis means, and the data provider builds the predictive model. . The shipping market forecast operation system according to claim 14 ,
Said predictive analysis means is characterized in that a predictive model building method is developed by a developer of said predictive analysis means, and said predictive model is constructed by a business operator other than said data provider. Market forecast operation system.

Images