GENERAL FIELD OF THE INVENTION

The present invention pertains to a method to control the culture of a next crop (i.e. the crop for a new growing season), the method comprising the provision of data

corresponding to the future market value of the said next crop by using a model to predict the market value of such crop, based on input parameters related to the growth of the crop, these parameters being chosen from genetic parameters and cultivating parameters of the crop. The invention also pertains to a computer program product programmed to automatically perform a method that results in the providing of said data, and to a method to produce a crop.

BACKGROUND ART

Over the past 30 years there has been a fairly steady increase in average yields of crop production, for example potato production, but corresponding revenues have not always shown a defined trend that correspond to the increase in yield, and may in some cases even trended downwardly over this period. Both acreage planted and yield have exhibited large fluctuations from year to year. Thus, expectations of total crop production and revenue are highly speculative until after the crop has been harvested. Therefore in the art a mechanism to forecast the expected level of production, and also to forecast the market value of the resulting mass of crop with a reasonable degree of accuracy has found to be a valuable tool for economic analysis and planning of crop cultivation. The market value in this respect can be any numerical or other figure representative for the economic value of the crop. Such value may depend on various factors such as nutritional composition of the crop, texture of the crop, resistance against decade after harvest, resistance against pathogens during growth, shape of the crop, size of the crop, an effect on the health of consumers of the crop, properties that influence processing (for example cooking) of the crop, etc. In the art various econometric models to predict production using selected data available well before the crop is harvested, in particular chosen from genetic parameters and cultivating parameters for the crop (for example for potatoes), are available for this cause.

However, it appears that the models, despite the level of sophistication still provide predicted market values that deviate too far from the actual market value at the end of the season in order to the producers to rely upon for planning their activities and business.

OBJECT OF THE INVENTION

It is an object of the invention to provide for an improved method to control the culture of a crop, in particular a method that is based on a better estimate at the start and during a culturing season for the actual market value of the crop. This way, the producers may be better able to control their activities and hence their business.

SUMMARY OF THE INVENTION

In order to meet the object of the invention, a method as outlined in the "General Field Of The Invention" section here above has been devised, wherein the method comprises for a previous culture of the crop, by using the model, the provision of data

corresponding to the future market value of this previous crop, wherein after the previous crop is harvested, its actual market value is compared with the predicted future market value to provide a difference between these values, wherein the said difference is used as an input parameter in the model to provide the data corresponding to the future market value of the said next crop, this value being used to control the culture of the said next crop.

Until this date, based on the general understanding that genetic parameters (i.e.

parameters related to the breed, i.e. the specific plant variety) and cultivating

parameters (i.e. parameters related to the cultivating methods such as when to plant, how to plant, how to fertilize, how to farm the ground etc.) are key for determining market value of a harvest of any crop, any fault in the prediction of the actual market value was believed to be due to using not completely correct values for either the genetic parameters or the cultivating parameters or anomalies in the model. Since genetic and cultivating parameters have been based on ages of tradition, in most cases the model is simply improved by fine-tuning the actual input parameters to decrease the difference between a predicted value and the actual market value, in order to hope in automatically providing a better prediction for any future harvest of any crop.

The current inventors however recognised that fine-tuning the parameters and model afterwards, to arrive at a match between predicted value and market value for a particular harvest of the past, is far from a guarantee that for the next culture of a crop the prediction of the market value will be more accurate. The prior art strategy has led to complicated models that provide a good match between predicted market value and actual market value for all harvests of crops in the past, but are still not capable of providing a good prediction for the future. As the current inventors recognised, there are other parameters, either known or unknown, that have an influence on the market value, but which parameters are not taken into account in the current models, simply because the parameter is not known, or it is not known how to fit such a parameter in a traditional econometric model. Instead of trying to unveil all of those parameters and fitting them into the existing models, the inventors recognised that it is adequate to use the apparent difference between the predicted value of the culture of a crop and its actual market value as an input parameter in the model to predict a next market value for a following culture of the crop. This way, all other parameters, either known or unknown can be taken into account in just one go without having the need to actually know them, let alone know what their value should be in order to be able and use it in a model. Indeed, this new way of predicting the market value of a culture of a crop presupposes that these unknown parameters do not change substantially between two consecutive cultures of crop. This is a reasonable assumption since most probably many parameters have an influence on the market value of crop (cost of transport, international trade restrictions, harvest of competitor crops, advertising activities, etc.), such that random variations may be reasonably balanced in between two consecutive cultures. Trends, leading away from actual balance, will still be found by taken into account results of consecutive harvest analysis.

The inventors recognised that the current invention is also potentially applicable to animal products, in particular edible products derived from production animals such as swine, bovine, poultry and fish. The invention also pertains to a computer program product (such as software, either being present on a mobile carrier such as a handheld device, or being run on a remote server system) programmed to automatically perform a method that results in the providing of data corresponding to a future market value of a crop by using a model as defined here above and in any of the appending claims.

The invention also pertains to a method to produce a crop comprising making a choice for genetic and cultivating parameters for the crop, based on a future market value of the crop using a model as defined here above and in any of the appending claims, and thereafter cultivating the crop according to the choices made.

EMBODIMENTS OF THE INVENTION

In a first embodiment of the invention the culture of the next crop is controlled by varying one or more of the genetic and cultivating parameters for the said next crop. In another embodiment the culture of the next crop is controlled by varying one or more of the following parameters for the crop: plant variety (a genetic parameter), or an cultivating parameter such as geographical location for culturing the crop, type of soil, method of farming (mechanically processing and feeding nutrients and water to) the ground, type of fertilizer, method of fertilizing the ground (globally, locally, according to a predetermined pattern, etc.), type of pesticide (herbicide and/or insecticide), method of using the pesticide (how much to use per surface area, how to apply, apply overall or locally, for example according to a pattern, etc.), growing season.

In yet another embodiment of the method according to the invention, in addition to the said difference, numeric market related data is used as input parameter for the model. Numeric market related data such as the number of advertorials for a type of crop, or the price trend for a type of crop, or the amount of a type of crop used in markets different than food etc. is relatively easy extracted from existing sources of information and is related to the actual market value of a crop.

In still another embodiment, in addition to the said difference, non-numeric market related data is used as input parameter for the model. Non-numeric market related data has never been actually assessed for use in an econometric model to predict the value of a culture of a particular crop (for example potatoes). However, it was inventor's recognition that such parameters may have a significant influence on the market value of a culture of that crop. For example, if a famous TV Chef regularly uses a type of crop in his recipes, this will have a positive impact on the market value. Also, the use of a crop or description of their use in popular TV shows, films or books may have an influence on market value. Although these data are typically not easy to catch in a representative number, they will have influence on the market value of a crop.

In order to be able and take this kind of data into account, in a further embodiment the non-numeric data is analysed using cognitive analytics. Cognitive analytics refers to a range of different analytical strategies that are used to learn about certain types of business related functions, based on analysis of numeric and non-numeric data. Certain types of cognitive analytics also may be known as predictive analytics, where data mining and other cognitive uses of data can lead to predictions for business intelligence (Bl). The general concept here is that large amounts of data from very diverse sources is collected and that specific software programs or other technologies analyze these data in depth to provide specific results that help a business get a better view of its own internal processes, how the market receives its products and services, customer preferences, how customer loyalty is generated or other key questions where accurate answers are used to provide a business with a competitive edge. In the art of crop culturing this is believed to lead to a better prediction of market value. Cognitive analytics is the application of these technologies to enhance human decisions. It takes advantage of cognitive computing's vast data-processing power and adds channels for data collection (such as sensing applications) and environmental context to provide practical business insights.

In yet another embodiment, during a growing season of the crop, dynamic data are used as additional input parameter for the model. Applicant recognised that next to static data, data that are set at the beginning of the growing season, it is advantageous to use dynamic data such as actual growth data, data regarding the market, weather data etc. These are used as (iterative) input data to arrive at a better prediction of data corresponding to the actual market value of the crop, such as yield and quality of the crop. In a next embodiment the market value is predicted using Big Data Analytics. Big Data Analytics is the process of examining large data sets containing a variety of data types - - i.e., big data - to uncover hidden patterns, unknown correlations, market trends, customer preferences and other useful business information (see i.a.

http://www.ibmbigdataanalytics.com). The primary goal of big data analytics is to help companies make more informed business decisions by enabling data scientists, predictive modellers and other analytics professionals to analyse large volumes of transaction data, as well as other forms of data that may be untapped by conventional business intelligence (Bl) programs. That could include Web server logs and Internet clickstream data, social media content and social network activity reports, text from customer emails and survey responses, mobile-phone call detail records and machine data captured by sensors connected to the Internet of Things. Big Data is not exclusively associated with semi-structured and unstructured data of that sort, on the contrary transactions and other structured data are also commonly considered to be valid components of big data analytics applications.

Big data can be analysed with the software tools commonly used as part of advanced analytics disciplines such as predictive analytics, data mining, text analytics and statistical analysis. Mainstream Bl software and data visualization tools can also play a role in the analysis process. But the semi-structured and unstructured data may not fit well in traditional data warehouses based on relational databases. Furthermore, data warehouses may not be able to handle the processing demands posed by sets of big data that need to be updated frequently or even continually - for example, real-time data on the growth performance of a crop (e.g. potatoes), weather information, energy prices, sudden trade restrictions or opportunities may all influence the actual market value at the end of the growing season. As a result, it is advantageous to turn to a newer class of technologies that includes Hadoop and related tools such as YARN, MapReduce, Spark, Hive and Pig as well as NoSQL databases. Those technologies form the core of an open source software framework that supports the processing of large and diverse data sets across clustered systems.

The use of Big Data Analytics enables a reasoned sampling of all information available that has some sort of influence of the market price of a crop (e.g. potatoes). It is recognised by the current inventors that using Big Data Analytics it is not needed to look at the full data to draw certain conclusions about the properties of the data: a sample may be good enough. Sampling (statistics) enables the selection of right data points from within the larger data set to estimate the characteristics of the whole population. For example, there are about 600 million tweets produced every day. Is it necessary to look at all of them to determine the topics that are discussed during the day? Is it necessary to look at all the tweets to determine the sentiment on each of the topics? In manufacturing different types of sensory data such as acoustics, vibration, pressure, current, voltage and controller data are available at short time intervals. To predict down-time it may not be necessary to look at all the data but a sample may be sufficient. The same is true for data related to the market value of any crop. Not every article about growing such crop or every TV cooking show wherein the crop is used needs to be viewed in full to extract proper input for the model.

In yet another embodiment of the method according to the invention, in order to minimize the difference between the (next) predicted market value and the actual market value of the new culture of crop, for example potatoes, the model as an output parameter delivers a proposal to change the genetic and/or cultivating parameter for the second culture of that crop (for example potatoes). It was applicant's recognition that not only the current method can lead to a better prediction of the actual market value of a harvest of a crop at the end of the growing season, it can also be used to propose to change the type of crop and/or growing method in order to arrive at a maximum market value. So for example, instead of a producer trying to adapt his expenses to the predicted market value, he will now get options to maximize this value at the end of the season. Such options may for example be to cultivate a different plant variety or to adapt the used cultivation technique.

The current invention enables new methods of cultivating a particular crop (e.g.

potatoes), i.a. a method to produce that crop comprising making a choice for a type of crop and a particular cultivation method, based on a predicted market value of the produced crop using any method as described here above and thereafter cultivating the chosen type of crop with the chosen method of cultivation.

The invention also enables a method to produce a crop, such as potatoes, comprising using Big Data Analytics to optimize the choice of type of crop (plant variety) and/or cultivation method, applying the chosen type of crop and/or cultivation method, cultivating the crop using the type of crop and/or the said cultivation method, and harvesting the crop. The invention will now be further explained using the following figures and example.

EXAMPLES

Figure 1 schematically depicts a system for applying the teachings of the present invention.

Figure 2 is an outline of a model for use in a method according to the present invention. Figure 3 is another outline of a model for use in a method according to the invention. Figure 4 provides a schematic representation of the spread in prediction and result.

Example 1 describes a method to provide pre-season and in-season predictions using the model of the current invention.

Figure 1

The core of the scheme is the third bar ("Analysis", indicated with reference numeral 12), which depicts the analysis block of the used data. This block, in this particular example incorporates Big Data platform and tools, as known from the art, Business

Intelligence (Bl) tools, statistical tools and geographical (GIS) tools. Together, they form the heart of the model, to provide the required output such as a predicted market value for a culture of crop such as potatoes, the proposed crop variety to be chosen or the proposed cultivation method. Such information may be communicated, as indicated in the lowermost block ("Output to users", indicated with reference numeral 13) to the producer (farmers and breeders) itself, but also to planners and agronomists, advisors and policy makers, and scientist and students etc.

The input for the analytical methods in this case is data present in an existing database for monitoring breeding and culturing of a crop ("Data structures", indicated with reference numeral 1 1)", which database is filled over the years i.a. with millions of data points regarding crop varieties, cultivation output, diseases, etc. derived from individual parties "Samples from crop"). These data in particular include, as indicated in the top block ("Data sources", numeral 10), data about the various lands used for cultivating the crop ("Lots, parcels and soil records"), data regarding varieties ("Cultivars and varieties") and data corresponding to various individual producers ("Farming, treatments and growing actions") as well as weather related data ("Weather and climate") and any other potentially relevant data such as images of parcels ("Images and other data"). These are data that have been traditionally gathered in the art of crop culturing, including methods to obtain new or improved plant varieties. A novel aspect is that other data, either numeric or non-numeric can now also be taken into account in the analytics by as direct input.

Figure 2

Figure 2 is an outline of a basic model 1 for use in a method according to the present invention. In this model 1 , various data 2, corresponding to traditional input parameters (breed, climate, traits, soil, farming goal, etc.) for econometric models for the provision of data corresponding to the future market value of the crop are used, in a first modelling step, in this case particular case to predict (indicated with numeral 3) the actual market value of a harvest of that crop (e.g. potatoes). This market value $P is indicated with reference numeral 4. This predicted market value is compared with the actual market value $M, indicated with reference numeral 5, and a value corresponding to this difference is used as another input variable, via line 6, in a next prediction cycle, for example to account for the influence of non-traditional data (i.e. data not equal to plant variety and cultivating method related data).

Figure 3

Figure 3 is an outline of a slightly more sophisticated model 1 ' (when compared to the model as depicted in figure 2) for use in a method according to the present invention. Also in this model 1 ', various data 2, corresponding to traditional input parameters for econometric models are used to provide data corresponding to the future market value of the crop. Corresponding to the model of figure 2, this data, indicated with numeral 4'is provided as output (via line 3), and compared with actual market data 5'. The difference is used, via line 6, as input for the model.

In this more sophisticated model iterative input parameters 7 (such as field

measurements during growth of the crop, images of the crop, weather data, contents, quality measurement data, market data etc.) are also used as input for the model to arrive at a better prediction. These are so-called in-season data that cannot be provided before the growing season (in contrast to the static data 2 which are provided only as pre-season data). Also the settings for yield and quality may be a variable that is taken into account as indicated with line 8. If the settings are mere pre-season data, no adaptation during the growing season is necessary and line 8 may be dispensed with. However, it is foreseen that settings may vary during the growing season and thus, the model may be optimized by taking this into account.

Figure 4

Figure 4 provides a schematic representation of the spread in prediction and result. The x-axis represents time that passes during the growing season. The y-axis represents the yield of the crop, i.e. information corresponding to the (future) market value of the harvested crop. At the beginning of the season, the maximum predicted yield "a" and the minimum predicted yield "b" represent a significant spread in yield. During the growing season, using iterative input parameters and optionally moving targets for quality the predicted yield "c" evolves. At the end of the season, the predicted yield c, when using the current model is the same as the new prediction for a crop of the next growing season, indicated in the figure by the circle formed between "a" and "b" coinciding with c. As can be seen, by using the current model, and thus taking into account dynamic data such as in-season data and the difference between predicted market value and actual market value, the model continuously provides improved output.

Example 1

Example 1 describes a method to provide pre-season and in-season predictions using the model of the current invention. This example describes a method to provide preseason and in-season yield and quality predictions in order to control the culture of a crop. The method has been developed to derive and construct a predictive model for yield and defined quality metrics (e.g. size/shape, specific weight, content of the crop etc.) based on identified input parameters. A list of input parameters is part of the developed model. The predictive capability of the model can be configured to allow both pre-season prediction (based on location, soil type, sowing date, cultivar type/variety, and local climate) and in-season dynamic (iterative) prediction. Historic, structured data is input to start the initial model. The model is self-learning based on iterative data inputs, i.a. using the difference between actual market value and predicted market value of a previous harvest of crop, and will thus be improved in time. The method uses a blending of pure data-drive models and classic deterministic models with relative weight factors and parameters calculated via machine learning algorithms.

Data inputs for the model in a first category are so called "Field growth and development data" that may be derived from several or many sites, from many cultivars, traits and varieties. Typically multiple samples per site at different growth stages of parameters are used such as soil analysis, growth stage dates, above and below ground biomass, height, ripening etc. all depending on the type of crop. Another category may be formed by "Field photos", for example a series of site photos, satellite picture or drone pictures of a field where the crop is growing. A third category may be "Weather data" such as current or a history of typical parameters such as temperature, humidity, wind force, hours of sun shine, etc. Yet another category may be formed by so-called "Crop quality data", optionally for all sites, farms, traits, varieties etc. Typically multiple samples are recorded per site with parameters such as (depending i.a. on the type of crop) grain composition data (content matter, protein, elements, etc.) and grain physical data (specific weight, size and shape, etc.).

Use of the model leads to a prediction for crop yield and defined quality metrics (e.g. grain size/shape, content matter), based on the identified input parameters. The predictive capability of the model allows both pre-season planning and prediction (based on location, soil type, sowing date, cultivar type/variety, local climate) and in-season dynamic (iterative) prediction (pre-season inputs plus identified in-season crop growth and development metrics and localized weather prediction models). Two key outputs may be provided with the model, viz. pre-season crop yield and quality predictions based on region and associated soil type and climate e.g. to predict the best cultivar type/variety by region for particular yield/quality outcomes, and dynamic (iterative) predictions within the growing season (i.a. based on crop growth and development inputs). Advantages of deploying this method and using the developed models are that one is able to arrive at better economic predictions and thus, it will become easier to control the actual culture of a crop. Better choice of the mass of crop to aim at, the production location and time, etc. This may lead to less water usage, less usage of crop protection materials, less usage of fertilizers and better farming rotation planning. As crop yield per area increases, less land is needed for the same production levels.

In life stock farming, very similar data sources may be used as input for the model. Trait and variety properties, content of feed, breeding and farming targets, growth stages, external measurements, weights, height, length of all body properties, body temperature, heart beat and other parameters will be input and through data analysis, initial estimates can be derived for model parameters. Likewise classical animal models will be mixed and the relative weight of the different models will be calculated based on measured growth rates and other factors. Likewise this may lead to savings in feed, water usage, use of animal drugs, like antibiotics and other input. The constructed model will improve with additional and iterative input because of learning capabilities the model provides.