WO2007089184A1

WO2007089184A1 - Dairy farm decision support system

Info

Publication number: WO2007089184A1
Application number: PCT/SE2006/050343
Authority: WO
Inventors: Bohao Liao
Original assignee: Delaval Holding Ab
Priority date: 2006-01-31
Filing date: 2006-09-19
Publication date: 2007-08-09

Abstract

A dairy farm management support system includes an input interface (130), a rules engine (110), a decision engine (120) and an output interface (140) for presenting at least one proposed dairy farm decision to a human user. The rules engine (110) receives a first set of input parameters (Drand), which reflect a decision basis for a particular type of dairy farm decisions. Via the input interface (130), the rules engine (110) also receives user generated input data (Decu), which represent a number of manual decisions in respect of the first set of input parameters (Drand). In response to the first set of input parameters (Drand) and the accompanying user-generated input data (Decu), the rules engine (110) generates a set of decision rules (RDec). Then, the decision engine (120) receives dairy farm data (D) reflecting animal parameters of the same type as the first set of input parameters (Drand), applies the set of decision rules (RDec) to the dairy farm data (D) and derives at least one proposed dairy farm decision of the particular type, which is presented via the output interface (140).

Description

Dai ry Farm Decision Support System

THE BACKGROUND OF THE INVENTION AND PRIOR ART

The present invention relates generally to support solutions intended to facilitate a dairy farmer's decision making in respect of his/her herd of livestock. More particularly the invention relates to a system according to claim 1 and a corresponding me- thod according to claim 15. The invention also relates to a computer program product according to claim 28 and a computer readable medium according to claim 29.

Dairy farm management decision-making is generally a complex task, which is difficult to model mathematically with a satisfying degree of accuracy. Namely, in many situations, the rules underlying a particular decision are not well defined, or cannot be formulated in concrete terms by the farmer. Instead, these types of decisions may be highly influenced by the farmer's intuition, or tacit experience. For example, even if it normally is appropriate to inseminate a cow when she is approximately 60 days in milk, this is not a rigid rule. On the contrary, many other factors may have impact on an actual insemination decision for a particular animal. However, often the farmer himself/herself is not explicitly aware of exactly which these factors are, or how they influence the insemination decision. Therefore, it may be practically impossible to formulate a Boolean, or other mathematical model, for the decision process.

Moreover, even if such a model could be formulated it would typically only be applicable to a particular farm or region. Other farms, regions or countries might require completely different models to satisfy the needs of the farmers who are active there.

The published European patent application 657 098 reveals an information system for automatically obtaining various data con- cerning animals in dairy farming, e.g. milk flow characteristics, reproductive status and/or health status. Here, parameter values are registered either automatically or manually. Based on the registered data expected future parameter values are determined, and an alarm is generated if the difference between the expec- ted value and an actual value is greater than a permissible deviation. In case of such an alarm, the farmer can investigate a possible cause of the alarm, and determine appropriate measures.

U.S. Patent No. 6,405,672 discloses a similar system for moni- toring the physical condition of a herd of livestock. Here, a number of animal-related parameters are automatically assessed for each animal individually. Based on historical data, future parameters are predicted, and if an error between a predicted value and a corresponding measured value falls outside of an automa- tically generated confidence interval, an alarm is produced indicating a potential unhealthy condition for the animal in question.

Although the above systems indeed may facilitate the farmer's supervision of his/her herd and its health status, none of the solutions is capable of providing assistance in the farm manage- ment decision process as such.

SUMMARY OF THE INVENTION

The object of the present invention is therefore to provide a solution, which alleviates the above problems and thus offers explicit decision support in the process of managing a dairy farm herd of livestock.

According to one aspect of the invention, the object is achieved by the system described initially, wherein the system includes a rules engine and a decision engine. The rules engine is adapted to receive a first set of input parameters reflecting a decision basis for a particular type of dairy farm decisions, receive user- generated input data representing a number of manual decisions in respect of the first set of input parameters, and generate a set of decision rules based on the first set of input parameters and the user-generated input data. The decision engine is adapted to receive dairy farm data reflecting animal parameters of the same type as the first set of input parameters, apply the set of decision rules to the dairy farm data, derive at least one proposed dairy farm decision of the particular type, and forward output data reflecting the at least one proposed dairy farm decision to the output interface for presentation to a human user.

This system is advantageous because it is capable of modeling decision processes that cannot be explicitly formulated. Hence, also in very complex farm management decision making the system can provide valuable support to the farmer.

According to one preferred embodiment of this aspect of the in- vention, the system includes a random data generator, which is adapted to produce the first set of input parameters. These parameters include a selection of parameter types, and for each parameter type a number of exemplary values are generated. The parameter types are selected based on the particular type of dairy farm decisions, and the exemplary values are distributed (e.g. according to a normal distribution) between a minimum value and a maximum value, which in turn, likewise are adapted to the particular type of dairy farm decisions. Thereby, appropriate training data is provided based on which the farmer can illustrate to the system how he/she wishes that decisions of a certain kind should be made. Consequently, the system becomes prepared for similar future decision making situations in which it can provide support.

According to another preferred embodiment of this aspect of the invention, the rules engine includes an artificial neural network, which is adapted to be trained based on the first set of input parameters and the user-generated input data. After adequate training, the resulting trained artificial neural network represents the set of decision rules to be applied by the rules engine, which thus likewise is an artificial neural network. Alternatively, the rules engine includes a processing unit, which is adapted to generate the set of decision rules based on linear correlation computations between the first set of input parameters and the user-ge- nerated input data. The results of these correlation computations reflect the set of decision rules, which are applied by the rules engine. Both these strategies are desirable in that they are highly flexible and can be used to support decisions over a wide range of dairy farm scenarios.

According to yet another preferred embodiment of this aspect of the invention, the system includes a data collection unit. This unit is adapted to receive a number of signals registered by at least one animal-related sensor means, and as a result produce at least a sub-set of the dairy farm data. Hence, large amounts of data may be collected automatically to provide a solid decision basis.

According to still another preferred embodiment of this aspect of the invention, the input interface is adapted to receive manually entered animal parameters, and forward these parameters to the decision engine to represent at least a sub-set of the dairy farm data. Thereby, the decision basis may either be supplemented manually, or the entire decision basis may be entered manually.

According to a further preferred embodiment of this aspect of the invention, the input interface is adapted to receive manually entered effect data reflecting the outcome of one or more actual farm management decisions made in agreement with the at least one proposed dairy farm decision. The input interface is here also adapted to forward the effect data to the rules engine, so that the set of decision rules represent an improved support qua lity. Specifically, the rules engine is adapted to modify the set of decision rules based on the effect data and the farm data, such that at least one effect of an actual farm management decision made in agreement with any future proposed dairy farm decision generated based on the modified decision rules is expected to attain an effect, which is superior to the effect of a corresponding actual farm management decision made in agreement with a proposed dairy farm decision resulting from the unmodified decision rules.

According to other preferred embodiments of this aspect of the invention, the proposed dairy farm decisions relate to inseminations, dry-cow states, determining individual animal energy demands, ketosis testing and/or mastitis testing.

In the case of insemination decision support the dairy farm data includes a first input parameter reflecting an animal identity. Additionally, the dairy farm data here includes parameters, which for each animal identity reflect a respective number of days in milk, a peak milk yield in a latest lactation, a conception rate of any previous lactations (i.e. the average number of insemina- tions required for the previous conceptions), a heat-detection status, a time period since a last heat date, a number of lactation periods, a body condition scoring, whether the animal has been selected for culling, and/or an activity status. The output data produced by the decision engine includes a first output parame- ter indicating the animal identity, and a second output parameter reflecting whether the animal having this identity should be inseminated within a particular time period.

In the case of decision support in respect of dry-cow states the dairy farm data includes a first input parameter reflecting an animal identity, and one or more parameters, which for each animal identity reflect a body condition scoring, a number of days in milk, a number of lactation periods, whether the animal has been selected for culling, expected calving data, mastitis event data of any previous lactation, ketosis event data of any previous lactation. The output data produced by the decision engine includes a first output parameter indicating the animal identity, and a second output parameter reflecting whether the animal having this identity should initiate a dry-off period within a particular time period or a dry-off treatment should be initiated within a particular time period.

In the case the decision support relates to determination of individual animal energy demands, the dairy farm data includes a first input parameter reflecting an animal identity, and one or more of parameters, which for each animal identity reflect a number of days in milk, a breeding status, a body condition scoring, a body-condition-scoring trend, a latest milk yield, a milk- yield trend, a number of lactation periods and a fat and protein concentration of the milk produced by the animal. The output data produced by the decision engine includes a first output parameter indicating the animal identity, and a second output parameter reflecting whether the animal with this identity within a particular time period is expected to have an unaltered, an increased or a decreased energy demand.

In the case the decision support relates to ketosis testing, the dairy farm data includes a first input parameter reflecting an animal identity, and one or more of parameters, which for each animal identity reflect a beta-hydroxybutyrate value, a number of days in milk, a body condition scoring, a body-condition-scoring trend (positive or negative), a latest milk yield and a milk-yield trend (positive or negative). The output data produced by the decision engine includes a first output parameter indicating the animal identity, and a second output parameter reflecting whether the animal with this identity should be tested for ketosis within a particular time period.

In the case the decision support relates to mastitis testing, the dairy farm data includes a first input parameter reflecting an animal identity, and one or more of parameters, which for each animal identity reflect a status of a low-activity alarm, a milk con ductivity, a relative milk yield, a somatic cell count and a number of days in milk. The output data produced by the decision engine includes a first output parameter indicating the animal identity, and a second output parameter reflecting whether the animal with this identity should be tested for mastitis within a particular time period.

According to another aspect of the invention the object is achieved by the method described initially, involving the steps of receiving a first set of input parameters reflecting a decision basis for a particular type of dairy farm decisions; receiving user-generated input data representing a number of manual decisions in respect of the first set of input parameters; generating a set of decision rules based on the first set of input parameters and the user-generated input data; receiving dairy farm data reflecting animal parameters of the same type as the first set of input parameters; applying the set of decision rules to the dairy farm data; deriving at least one proposed dairy farm decision of the particular type; and presenting output data reflecting the at least one proposed dairy farm decision on said format.

The advantages of this method, as well as the preferred embodiments thereof, are apparent from the discussion hereinabove with reference to the proposed system.

According to a further aspect of the invention the object is achieved by a computer program product, which is directly loadable into the internal memory of a computer, and includes software for controlling the above proposed method when said program is run on a computer.

According to another aspect of the invention the object is achieved by a computer readable medium, having a program recorded thereon, where the program is to control a computer to perform the above-proposed method.

Further advantages, advantageous features and applications of the present invention will be apparent from the following des cription and the dependent claims.

BRIEF DESCRI PTION OF THE DRAWINGS

The present invention is now to be explained more closely by means of preferred embodiments, which are disclosed as examples, and with reference to the attached drawings.

Figure 1 shows a block diagram over a system according to one embodiment of the invention;

Figure 2 illustrates, by means of a first flow diagram, a general method of controlling a computer apparatus according to the invention when producing a set of decision rules; and

Figure 3 illustrates, by means of a second flow diagram, a general method of controlling a computer apparatus according to the invention when applying the decision rules to derive proposed decisions.

DESCRIPTION OF PREFERRED EMBODIMENTS OF THE INVENTION

We refer initially to figure 1 , which shows a block diagram over a system for dairy farm management support according to one em- bodiment of the invention. The system includes an input interface 130, an output interface 140, a rules engine 1 10 and a decision engine 120.

According to one preferred embodiment of the invention, the system also includes a random data generator 1 15. This unit is adapted to produce a first set of input parameters D_rand representing training data for a particular type of dairy farm decisions. The parameters D_rand include a selection of parameter types, and for each parameter type the random data generator 1 15 produces a number of exemplary values.

The parameter types are selected to be relevant for the parti cular type of dairy farm decisions, i.e. the parameter types reflect factors that are important to consider when making the decision in question. The parameter types may either be proposed automatically by the system, or they may be selected manually by the farmer/user. In any case, the system is preferably adapted to allow the farmer/user to modify the parameter types at any stage of the process.

Such interaction is accomplished via the input and output interfaces 130 and 140 respectively. The input interface 130 is adap- ted to receive user-generated input data Decu, and may thus be represented by a keyboard, a touch screen, a pointer device (e.g. a so-called computer mouse) or a voice recognition interface, or arbitrary combination thereof. The output interface 140 is adapted to present automatically generated output data Dec_P to a human user. Consequently, this interface preferably includes a computer display. However, a printer and/or an acoustic interface are equally well conceivable.

The exemplary values for each parameter type of the parameters in D_rand are selected such that the values are distributed between a minimum value and a maximum value. The minimum and maximum values respectively, in turn, are representative for the particular type of dairy farm decisions for which support is desired. In order to attain a good data quality, it is further preferable if the values are distributed according to a normal, or Gaussian, distribution. However, in many cases, other types of distributions may be more appropriate.

The rules engine 1 10 is adapted to receive a first set of input parameters D_rand, preferably from the random data generator 1 15. However, according to the invention, any other type of data source is conceivable, e.g. a storage medium, a remote resource accessible via a computer network, or manual entry. The first set of input parameters D_rand constitute a decision basis for a particular type of dairy farm decisions, e.g. with respect to insemination timing. The below table exemplifies the first set of input parameters D_rand being applicable to insemination decisions. For each set of parameter values, the rightmost column shows examples of manual (i.e. human made) decisions by a particular farmer/user. Hence, this column represents user-generated input data Decu-

In the above table, the first parameter indicates the number of days in milk (DIM) for the animal, the second parameter indicates the peak yield of milk (here in liters), the third parameter shows the average number of inseminations that have been required for the animal's previous pregnancies, the fourth parameter indicates whether the animal is in heat presently, the fifth parameter designates the number of days since the latest heat, the sixth parameter specifies the number of previous lactations and the seventh parameter represents a body condition scoring (BCS) for the animal, which is either manually assigned, or automatically assessed, e.g. according to imaging system described in the pub lished international patent application WO 2004/012146.

According to the proposed invention, the rules engine 1 10 receives the first set of input parameters D_rand along with the user- generated input data DeCu, which represent the farmer's manual decisions in respect of these first parameters D_rand, i.e. in case of insemination decisions, whether a particular cow should be inseminated within a particular time period or not. Based on the first set of input parameters D_rand and the user-generated input data Decu, the rules engine 1 10 generates a set of decision rules R_Dec.

These rules Ro_ec, in turn, are fed to the decision engine 120. Thereby, the decision engine 120 becomes adapted to generate future decision proposals in line with the decision strategies reflected by the user-generated input data Decu (i.e. the exemp- lifying decisions entered by the farmer in connection with the first set of input parameters D_rand).

When operating the system in a decision-support mode after the above-described training mode, the system receives dairy farm data D, which reflect animal parameters of the same type as the first set of input parameters D_rand. Thus, if the system has been trained in respect of insemination by means of the above first set of input parameters D_rand, the dairy farm data D comprises parameters that for each animal indicate a respective number of days in milk, a peak yield of milk, an average number of inseminations previously required, whether the animal is in heat, a number of days since the latest heat, a number of previous lactations and a body condition scoring.

The decision engine 120 applies the set of decision rules R_Dec to the dairy farm data D and derives at least one proposed dairy farm decision of the particular type, e.g. whether the individual members of a group of cows should be inseminated within a particular time period or not. The decision engine 120 then forwards the at least one proposed dairy farm decision in the form of output data Dec_P to the output interface 140 for presentation to the farmer/user.

Continuing with the insemination example, the system may produce the following table in response to a farmer's/user's inquiry regarding which animals in his/her herd of livestock that should be inseminated today.

Hence, the output data Dec_P here includes a first output parameter reflecting the animal identity and a selection of input parameters that were used to reach the conclusion that insemina- tion is recommended. However, animals that should not be inseminated are not listed here. Of course, the output data Dec_P may equally well be organized such that instead all animal identities in the herd are included, and the second output parameter reflects whether a particular animal should be inseminated or not.

According to one preferred embodiment of the invention, the rules engine 1 10 and the decision engine 120 include a respective artificial neural network. This artificial neural network is first trained (or programmed) in the rules engine 1 10 based on the first set of input parameters D_rand and the user-generated input data Decu, such that the resulting trained artificial neural network represents the set of decision rules R_Dec- Thereafter, the trained artificial neural network (i.e. the set of decision rules R_Dec) is copied into the decision engine 120.

According to another preferred embodiment of the invention, the rules engine 1 10 and the decision engine 120 instead include a respective processing unit. The processing unit in the rules engine 1 10 is adapted to generate the set of decision rules R_Dec based on linear correlation computations between the first set of input parameters D_rand and the user-generated input data Decu- Analogously, the thus produced set of decision rules R_Dec is transferred to the processing unit in the decision engine 120, where the rules R_Dec are applied to dairy farm data D to derive at least one proposed dairy farm decision.

Preferably, the system also includes a data collection unit 135, which is connected to, by wire or wirelessly (e.g. by means of radio-, optical- or ultrasonic transmission resources), a number of animal-related sensor means, (not shown) i.e. sensors located on/in the animals, or at implements interrelating with the animals, such as milking apparatuses. Thereby, the data collection unit 135, may receive a number of signals d^ d₂,..., d_n registered by these sensor means, and based thereon produce at least a sub-set of the dairy farm data D. It is also desirable that the input interface 130 is adapted to receive manually entered animal parameters, and forward these parameters to the decision engine 120 to represent at least a sub-set of the dairy farm data D. Namely, thereby, the farmer/user can supplement and/or correct/modify any automatically registered data.

Specifically, according to one preferred embodiment of the invention, the input interface 130 is further adapted to receive manually entered effect data E_FB, which reflects the outcome of at least one actual farm management decision made in agreement with the at least one proposed dairy farm decision. This data E_FB are forwarded to the rules engine 1 10, which in turn, is adapted to modify its decision rules R_Dec based on the effect data E_FB and the corresponding farm data D (i.e. the data linked to effect data E_FB). The decision rules R_Dec are modified such that at least one effect of an actual farm management decision made in agreement with any future proposed dairy farm decision generated based on the modified decision rules R_Dec is expected to be improved, i.e. attain an effect superior to the effect of a corresponding actual farm management decision made in agreement with a proposed dairy farm decision generated based on the unmodified decision rules Ro_ec- Naturally, the modified decision rules R_Dec are also forwarded to the decision engine 120, so that the updated decision rules R_Dec gain influence over decision proposals in respect of future dairy farm data D.

Instead of, or as a complement to, the above, the proposed dairy farm decision may relate to a dry-cow state. In this case, the dairy farm data D includes a first input parameter reflecting an animal identity, and for each identity, parameters indicating one or more of the following: a body condition scoring, a number of days in milk, a lactation period number, whether the animal has been selected for culling, expected calving data for the animal, mastitis event data of any previous lactation for the animal and ketosis event data of any previous lactation for the animal. Here, the output data Dec_P preferably includes a first output parameter reflecting the animal identity, and a second output parameter reflecting whether the animal having this identity, within a particular time period, should initiate a dry-off period or a dry-off treatment should be initiated.

According to another embodiment of the invention, the proposed dairy farm decision relates to determination of individual animal energy demands. To this aim, the dairy farm data D includes a first input parameter reflecting an animal identity, and for each identity, parameters indicating one or more of the following: a number of days in milk, a breeding status (i.e. whether the animal is pregnant of not), a body condition scoring, a body-condition-scoring trend (positive or negative), a latest milk yield, a milk-yield trend (positive or negative), a lactation period number, and a fat and protein concentration of the milk produced by the animal. In this case, the output data Dec_P preferably includes a first output parameter reflecting the animal identity, and a second output parameter reflecting whether the animal with this identity is expected to have an unaltered, an increased or a de creased energy demand within a particular time period, say a following week.

According to another embodiment of the invention, the proposed dairy farm decision relates to whether one or more animals should be subjected to ketosis testing. Here, the dairy farm data D includes a first input parameter reflecting an animal identity, and for each identity, parameters indicating one or more of the following: a beta-hydroxybutyrate (BBH) value, a number of days in milk, a body condition scoring, body-condition-scoring trend (positive or negative), a latest milk yield and a milk-yield trend (positive or negative). Preferably, the output data Dec_P specifies a first output parameter reflecting the animal identity, and a second output parameter reflecting whether the animal with this identity should be tested for ketosis within a particular time pe- riod, e.g. today.

According to yet another embodiment of the invention, the proposed dairy farm decision relates to whether one or more animals should be subjected to mastitis testing. To this aim, the dairy farm data D includes a first input parameter reflecting an animal identity, and for each identity, parameters indicating one or more of the following: a status of a low-activity alarm (i.e. essentially whether a Registered activity level falls below a threshold level, which may depend on the animal's present movement space), a milk conductivity, a relative milk yield, a somatic cell count (SCC) and a number of days in milk. Preferably, the output data Deep specifies a first output parameter reflecting the animal identity, and a second output parameter reflecting whether the animal with this identity should be tested for mastitis within a particular time period, e.g. during the present week.

To sum up, we will now describe the general method of controlling a computer apparatus for generating the set of decision rules according to the invention with reference to the flow diagram in figure 2. An initial step 210 generates a first set of input parameters, preferably by a random algorithm, to constitute a decision basis for a particular type of dairy farm decisions. Then, a step 220 presents the first set of input parameters to a user, for instance in the form of a query set on a computer display. Subsequently, a step 230 receives user-generated input data in respect of the first set of input parameters, i.e. a number of manual decisions made on the basis of the circumstances defined by these parameters. Finally, a step 240 generates a set of decision rules based on the first set of input parameters and the user-generated input data. The set of decision rules here extrapolates a number of generic decision principles from the manual decisions of the step 230, such that these principles can be applied to future decision scenarios.

Turning now to the flow diagram in figure 3, we will describe the general method of controlling a computer apparatus for deriving dairy-farm-decision proposals. A first step 310 receives dairy farm data reflecting animal parameters of the same type as the first set of input parameters received in the step 210 above. A subsequent step 320 applies the set of decision rules generated in the step 240 to the dairy farm data, and derives at least one proposed dairy farm decision of the particular type mentioned above. A final step 330 then presents output data reflecting said at least one decision on a format, which is comprehensive to a human, e.g. visually on a graphical display and/or acoustically via a loudspeaker.

All of the process steps, as well as any sub-sequence of steps, described with reference to the figures 2 and 3 above may be controlled by means of a programmed computer apparatus. Moreover, although the embodiments of the invention described above with reference to the drawings comprise computer apparatus and processes performed in computer apparatus, the invention thus also extends to computer programs, particularly computer programs on or in a carrier, adapted for putting the invention into practice. The program may be in the form of source code, object code, a code intermediate source and object code such as in partially compiled form, or in any other form suitable for use in the implementation of the process according to the invention. The program may either be a part of an operating system, or be a separate application. The carrier may be any entity or device capable of carrying the program. For example, the carrier may comprise a storage medium, such as a Flash memory, a ROM (Read Only Memory), for example a DVD (Digital Video/Versatile Disk), a CD (Compact Disc) or a semi- conductor ROM, an EPROM (Erasable Programmable Read- OnIy Memory), an EEPROM (Electrically Erasable Programmable Read-Only Memory), or a magnetic recording medium, for example a floppy disc or hard disc. Further, the carrier may be a transmissible carrier such as an electrical or optical signal which may be conveyed via electrical or optical cable or by radio or by other means. When the program is embodied in a signal which may be conveyed directly by a cable or other device or means, the carrier may be constituted by such cable or device or means. Alternatively, the carrier may be an integrated circuit in which the program is embedded, the integrated circuit being adapted for performing, or for use in the performance of, the relevant processes.

The term "comprises/comprising" when used in this specification is taken to specify the presence of stated features, integers, steps or components. However, the term does not preclude the presence or addition of one or more additional features, integers, steps or components or groups thereof.

The invention is not restricted to the described embodiments in the figures, but may be varied freely within the scope of the claims.

Claims

1 . A dairy farm management support system comprising: an input interface (130) adapted to receive user-generated input data (Decu), and an output interface (140) adapted to present automatically generated output data (Dec_P) to a human user characterized in that the system further comprises: a rules engine (1 10) adapted to receive a first set of input parameters (D_rand) reflec- ting a decision basis for a particular type of dairy farm decisions, receive user-generated input data (Decu) representing a number of manual decisions in respect of the first set of input parameters (D_rand), and generate a set of decision rules (R_DΘC) based on the first set of input parameters (D_rand) and the user-generated input data (Decu), and a decision engine (120) adapted to receive dairy farm data (D) reflecting animal parame- ters of the same type as the first set of input parameters

( D rand) j apply the set of decision rules (R_DΘC) to the dairy farm data (D), derive at least one proposed dairy farm decision of the particular type, and forward output data (Dec_P) reflecting the at least one proposed dairy farm decision to the output interface (140).

2. The system according to claim 1 , characterized in that it comprises a random data generator (1 15) adapted to produce the first set of input parameters (D_rand), such that these parameters (D_rand) comprise a selection of parameter types and a number of exemplary values are generated for each parameter type, the parameter types being selected based on the particular type of dairy farm decisions, and the exemplary values being distri- buted between a minimum value and a maximum value which are adapted to the particular type of dairy farm decisions.

3. The system according to claim 2, characterized in that the exemplary values are distributed according to a normal distribution.

4. The system according to any one of the preceding claims, characterized i n that the rules engine (1 10) comprises an artificial neural network adapted to be trained based on the first set of input parameters (D_rand) and the user-generated input data (Decu), and generate the set of decision rules (R_Dec) in the form of a resulting trained artificial neural network.

5. The system according to any one of the claims 1 to 4, characterized in that the rules engine (1 10) comprises a processing unit adapted to generate the set of decision rules (Ro_ec) ba- sed on linear correlation computations between the first set of input parameters (D_rand) and the user-generated input data (Decu).

6. The system according to any one of the preceding claims, characterized in that it comprises a data collection unit (135) adapted to receive a number of signals (d^ d₂, d_n) registered by at least one animal-related sensor means, and produce at least a sub-set of the dairy farm data (D) based on the signals (d-, , d₂, d_n).

7. The system according to any one of the preceding claims, characterized i n that the input interface (130) is adapted to receive manually entered animal parameters, and forward these parameters to the decision engine (120) to represent at least a sub-set of the dairy farm data (D).

8. The system according to any one of the preceding claims, characterized in that the input interface (130) is adapted to receive manually entered effect data (E_FB) reflecting at least one outcome of at least one actual farm management decision made in agreement with the at least one proposed dairy farm decision, and forward the effect data (E_FB) to the rules engine (1 10).

9. The system according to claim 8, characterized in that the rules engine (1 10) is adapted to modify the set of decision rules (Ro_ec) based on the effect data (E_FB) and the farm data (D) such that at least one effect of an actual farm management deci- sion made in agreement with any future proposed dairy farm decision generated based on the modified decision rules (R_Dec) is expected to attain an effect which is superior to the effect of a corresponding actual farm management decision made in agreement with a proposed dairy farm decision generated based on the unmodified decision rules (R_Dec)-

10. The system according to any one of the preceding claims, characterized in that the at least one proposed dairy farm decision relates to insemination, and the dairy farm data (D) comprises a first input parameter reflecting an animal identity, and at least one of a second input parameter reflecting a number of days in milk for the animal having said identity, a third input parameter reflecting a peak milk yield in a latest lactation for the animal having said identity, a fourth input parameter reflecting a conception rate of any previous lactation for the animal having said identity, a fifth input parameter reflecting a heat-detection status for the animal having said identity, a sixth input parameter reflecting a time period since a last heat date for the animal having said identity, a seventh input parameter reflecting a lactation period number for the animal having said identity, an eighth input parameter reflecting a body condition scoring for the animal having said identity, a ninth input parameter reflecting whether the animal having said identity has been selected for culling, and a tenth input parameter reflecting an activity status for the animal having said identity, and the output data (Dec_P) comprises a first output parameter reflecting the animal identity, and a second output parameter reflecting whether the animal having said identity should be inseminated within a particular time period.

1 1 . The system according to any one of the preceding claims, characterized in that the at least one proposed dairy farm decision relates to a dry-cow state, and the dairy farm data (D) comprises a first input parameter reflecting an animal identity, and at least one of a second input parameter reflecting a body condition scoring for the animal having said identity, a third input parameter reflecting a number of days in milk for the animal having said identity, a fourth input parameter reflecting a lactation period number for the animal having said identity, a fifth input parameter reflecting whether the animal having said identity has been selected for culling, a sixth input parameter reflecting expected calving data, a seventh input parameter reflecting mastitis event data of any previous lactation for the animal having said identity, an eighth input parameter reflecting ketosis event da- ta of any previous lactation for the animal having said identity, and the output data (Dec_P) comprises a first output parameter reflecting the animal identity, and a second output parameter reflecting whether the animal having said identity within a particular time period a dry-off period should be initiated or a dry-off treatment should be initiated.

12. The system according to any one of the preceding claims, characterized in that the at least one proposed dairy farm decision relates to determining individual animal energy demands, and the dairy farm data (D) comprises a first input parameter reflecting an animal identity, and at least one of a second input parameter reflecting a number of days in milk for the animal having said identity, a third input parameter reflecting a breeding status for the animal having said identity a fourth input parameter reflecting a body condition scoring for the animal having said identity, a fifth input parameter reflecting a body-condition- scoring trend for the animal having said identity, a sixth input parameter reflecting a latest milk yield for the animal having said identity, a seventh input parameter reflecting a milk-yield trend for the animal having said identity, an eighth input parameter reflecting a lactation period number for the animal having said identity, and a ninth input parameter reflecting a fat and protein concentration of the milk produced by the animal having said identity, and the output data (Dec_P) comprises a first output parameter reflecting the animal identity, and a second output parameter reflecting whether the ani- mal having said identity within a particular time period is expected to have an unaltered, an increased or a decreased energy demand.

13. The system according to any one of the preceding claims, characterized in that the at least one proposed dairy farm de cision relates to ketosis testing, and the dairy farm data (D) comprises a first input parameter reflecting an animal identity, and at least one of a second input parameter reflecting a beta-hydroxy- butyrate value for the animal having said identity a third input parameter reflecting a number of days in milk for the animal having said identity, a fourth input parameter reflecting a body condition scoring for the animal having said identity, a fifth input parameter reflecting a body-condition- scoring trend for the animal having said identity, a sixth input parameter reflecting a latest milk yield for the animal having said identity, a seventh input parameter reflecting milk-yield trend for the animal having said identity, and the output data (Dec_P) comprises a first output parameter reflecting the animal identity, and a second output parameter reflecting whether the animal having said identity should be tested for ketosis within a particular time period.

14. The system according to any one of the preceding claims, characterized in that the at least one proposed dairy farm decision relates to mastitis testing, and the dairy farm data (D) comprises a first input parameter reflecting an animal identity, and at least one of a second input parameter reflecting a status of a low- activity alarm for the animal having said identity a third input parameter reflecting milk conductivity for the animal having said identity, a fourth input parameter reflecting a relative milk yield for the animal having said identity, a fifth input parameter reflecting a somatic cell count for the animal having said identity, a sixth parameter reflecting a number of days in milk for the animal having said identity, and the output data (Dec_P) comprises a first output parameter reflecting the animal identity, and a second output parameter reflecting whether the animal having said identity should be tested for mastitis within a particular time period.

15. A method of generating dairy farm management support, comprising: receiving user-generated input data (DeCu), and presenting automatically generated output data (Dec_P) on a format perceivable by a human user, characterized by receiving a first set of input parameters (D_rand) reflecting a decision basis for a particular type of dairy farm decisions, receiving user-generated input data (DeCu) representing a number of manual decisions in respect of the first set of input parameters (D_rand), generating a set of decision rules (Ro_ec) based on the first set of input parameters (D_rand) and the user-generated input data (Decu), receiving dairy farm data (D) reflecting animal parameters of the same type as the first set of input parameters (D_rand), applying the set of decision rules (R_Dec) to the dairy farm data (D), deriving at least one proposed dairy farm decision of the particular type, and presenting output data (Dec_P) reflecting the at least one proposed dairy farm decision on said format.

16. The method according to claim 15, characterized by pro- ducing the first set of input parameters (D_rand) according to a random algorithm, such that the first set of input parameters (D_rand) comprises a selection of parameter types and a number of exemplary values for each parameter type, the parameter types being selected based on the particular type of dairy farm de cisions, and the exemplary values being distributed between a minimum value and a maximum value which are adapted to the particular type of dairy farm decisions.

17. The method according to claim 16, characterized by the exemplary values being distributed according to a normal distribution.

18. The method according to any one of the claims 15 to 17, characterized by implementing the set of decision rules (R_Dec) in the form of a trained artificial neural network having been trained based on the first set of input parameters (D_rand) and the user-generated input data (Decu).

19. The method according to any one of the claims 15 to 17, characterized by generating the set of decision rules (R_Dec) based on linear correlation computations between the first set of in- put parameters (D_rand) and the user-generated input data (Decu).

20. The method according to any one of the claims 15 to 19, characterized by collecting a number of signals (d^ d₂, d_n) registered by at least one animal-related sensor means, and producing at least a sub-set of the dairy farm data (D) based on the signals (d-, , d₂, d_n).

21 . The method according to any one of the claims 15 to 20, characterized by receiving manually entered animal parameters to represent at least a sub-set of the dairy farm data (D).

22. The method according to any one of the claims 15 to 21 , characterized by receiving manually entered effect data (E_FB) reflecting at least one outcome of at least one actual farm management decision made in agreement with the at least one proposed dairy farm decision, and modifying the set of decision rules (Ro_ec) based on the effect data (E_FB) and the farm data (D) such that at least one effect of an actual farm management decision made in agreement with any future proposed dairy farm decision generated based on the modified decision rules (Ro_ec) is expected to attain an effect which is superior to the effect of a corresponding actual farm management decision made in agreement with a proposed dairy farm decision generated based on the unmodified decision rules

23. The method according to any one of the claims 15 to 22, characterized by the at least one proposed dairy farm decision relating to insemination, and the dairy farm data (D) comprising a first input parameter reflecting an animal identity, and at least one of a second input parameter reflecting a number of days in milk for the animal having said identity, a third input parameter reflecting a peak milk yield in a latest lactation for the animal having said identity, a fourth input parameter reflecting a conception rate of any previous lactation for the animal having said identity, a fifth input parameter reflecting a heat-detection status for the animal having said identity, a sixth input parameter reflecting a time period since a last heat date for the animal having said identity, a seventh input parameter reflecting a lactation period number for the animal having said identity, an eighth input parameter reflecting a body condition scoring for the animal having said identity, a ninth input parameter reflecting whether the animal having said identity has been selected for culling, and a tenth input parameter reflecting an activity status for the animal having said identity, and the output data (Dec_P) comprises a first output parameter reflecting the animal identity, and a second output parameter reflecting whether the animal having said identity should be inseminated within a particular time period.

24. The method according to any one of the claims 15 to 23, characterized by the at least one proposed dairy farm decision relating to a dry-cow state, and the dairy farm data (D) comprising a first input parameter reflecting an animal identity, and at least one of a second input parameter reflecting a body condition scoring for the animal having said identity, a third input parameter reflecting a number of days in milk for the animal having said identity, a fourth input parameter reflecting a lactation period number for the animal having said identity, a fifth input parameter reflecting whether the animal having said identity has been selected for culling, a sixth input parameter reflecting expected calving data, a seventh input parameter reflecting mastitis event data of any previous lactation for the animal having said identity, an eighth input parameter reflecting ketosis event data of any previous lactation for the animal having said identity, and the output data (Dec_P) comprises a first output parameter reflecting the animal identity, and a second output parameter reflecting whether for the animal having said identity within a particular time period a dry-off period should be initiated or a dry-off treatment should be initiated.

25. The method according to any one of the claims 15 to 24, characterized by the at least one proposed dairy farm decision relating to determining individual animal energy demands, and the dairy farm data (D) comprising a first input parameter reflecting an animal identity, and at least one of a second input parameter reflecting a number of days in milk for the animal having said identity, a third input parameter reflecting a breeding status for the animal having said identity a fourth input parameter reflecting a body condition scoring for the animal having said identity, a fifth input parameter reflecting a body-condition- scoring trend for the animal having said identity, a sixth input parameter reflecting a latest milk yield for the animal having said identity, a seventh input parameter reflecting a milk-yield trend for the animal having said identity, an eighth input parameter reflecting a lactation period number for the animal having said identity, and a ninth input parameter reflecting a fat and protein concentration of the milk produced by the animal having said identity, and the output data (Dec_P) comprises a first output parameter reflecting the animal identity, and a second output parameter reflecting whether the animal having said identity within a particular time period is expected to have an unaltered, an increased or a decreased energy demand.

26. The method according to any one of the preceding claims 15 to 25, characterized by the at least one proposed dairy farm decision relating to ketosis testing, and the dairy farm data (D) comprising a first input parameter reflecting an animal identity, and at least one of a second input parameter reflecting a beta-hydroxy- butyrate value for the animal having said identity a third input parameter reflecting a number of days in milk for the animal having said identity, a fourth input parameter reflecting a body condition scoring for the animal having said identity, a fifth input parameter reflecting a body-condition- scoring trend for the animal having said identity, a sixth input parameter reflecting a latest milk yield for the animal having said identity, a seventh input parameter reflecting milk-yield trend for the animal having said identity, and the output data (Dec_P) comprises a first output parameter reflecting the animal identity, and a second output parameter reflecting whether the animal having said identity should be tested for ketosis within a particular time period.

27. The system according to any one of the claims 15 to 26, characterized by the at least one proposed dairy farm decision relating to mastitis testing, and the dairy farm data (D) comprising a first input parameter reflecting an animal identity, and at least one of a second input parameter reflecting a status of a low- activity alarm for the animal having said identity a third input parameter reflecting milk conductivity for the animal having said identity, a fourth input parameter reflecting a relative milk yield for the animal having said identity, a fifth input parameter reflecting a somatic cell count for the animal having said identity, a sixth parameter reflecting a number of days in milk for the animal having said identity, and the output data (Dec_P) comprises a first output parameter reflecting the animal identity, and a second output parameter reflecting whether the animal having said identity should be tested for mastitis within a particular time period.

28. A computer program product directly loadable into the internal memory of a computer, comprising software for controlling the steps of any of the claims 15 to 27 when said program is run on the computer.

29. A computer readable medium, having a program recorded thereon, where the program is to make a computer control the steps of any of the claims 15 to 27.