9/3/2018 Department of Veterinary and Animal Sciences Department of - PDF document

9/3/2018 Department of Veterinary and Animal Sciences Department of Veterinary and Animal Sciences Outline The decision making process The role of models From registration to information, I Basic production monitoring as a source of information Key figures and their properties Anders Ringgaard Kristensen Interpretation of key figures Limitations of traditional production monitoring Slide 2 Department of Veterinary and Animal Sciences Focus of this lecture Making decisions In last lecture, we discussed: • Planning, classical approach Decision making is based on knowledge: In this lecture we direct our • General knowledge : What you can read in a textbook on attention towards the other animal nutrition, animal breeding, agricultural engineering side of the management etc. cycle: • Context specific knowledge : What relates directly to the unique decision problem. Examples: • Production monitoring • The milk yield of dairy cow No. 678 when considered for We shall, however, try to look culling. at it from a decision making • The estrus status of sow No. 345 when considered for perspective. insemination. • The current daily gain of the slaughter pigs in House 5 when considering whether or not to adjust protein contents of the diet. When knowledge is represented in a form that may be used directly as basis for a decision, we call it information. Slide 4 Department of Veterinary and Animal Sciences Department of Veterinary and Animal Sciences Information sources From registrations to information General knowledge: We refer to a collection (typically in a database) of registrations of the same kind as data. • Look in a textbook We don’t use data directly for decision making (huge • Ask an expert amounts of data). Context specific knowledge: Before we can use data we need to reduce it through • Obtained through registrations (observations) in the herd: some kind of processing. • Traditional registrations The resulting information is used for decision making • Test day milk yield, cow 567 (which again requires processing: optimization). • Litter size of sow 123 • Sensor based registrations • Conductivity or temperature of milk from AMS • Accelerations of a sow from a censor node in an ear tag • Computer vision (image analysis) Slide 5 Slide 6 1

9/3/2018 Department of Veterinary and Animal Sciences Department of Veterinary and Animal Sciences Advanced example: Hogthrob A simple example of the path Data: Test day milk yields Activity Measurements – in Group Housed Pen Processing I: Calculating cumulated yields for individual cows over a standardized period and afterwards calculating • Accelerometer fitted on neck collar the herd average. • Acceleration in 3 dimensions Information: Average milk yield in the herd. • Four measurements per second Processing II: Linear programming using the Simplex algorithm. • Transfer PC via Blue Tooth Decision: Least cost feed ration for the cows. • Gestation house and farrowing crate The path from test day milk yields to feed ration is not unique: Video Recordings • Both processing steps could be replaced by other methods. • Four cameras used as web cam • Choosing a wise processing of data into information is an important issue in herd management! Slide 7 Slide 8 Cécile Cornou, LC-2373, IPH, KVL Department of Veterinary and Animal Sciences The Farrowing House Farrowing Data Collected – Farrowing house Slide 10 Department of Veterinary and Animal Sciences Department of Veterinary and Animal Sciences 2 days before Farrowing Activity Classification – Farrowing Activity Classification – Farrowing farrowing day Feeding: 7.15, 12.00, 15.30 Feeding / Rooting / Nesting Active Active Lying side 1 Lying sternally Lying side 2 Lying side 1 Lying sternally Lying side 2 Slide 11 Slide 12 2

9/3/2018 Department of Veterinary and Animal Sciences Department of Veterinary and Animal Sciences Information retrieval – Farrowing (or heat) An advanced example of the path: Hogthrob Data: Accelerations of a sow measured 4 times per second in 3 dimensions Processing I: Online time series analysis of the acceleration data using Dynamic Linear Models Information: Sow in heat? (yes/no). [Example: farrow] Processing II: Dynamic programming. Decision: Inseminate/leave open/cull Notice the reduction in the dimensionality of the information (one binary variable) compared to the data! Slide 13 Slide 14 Department of Veterinary and Animal Sciences Department of Veterinary and Animal Sciences Uncertain information The decision making process: Summary Classical methods assume certainty. In real life, certain information hardly exists. The purpose of monitoring is to improve the decisions However, the degree of uncertainty varies. Processing of registrations into information is necessary In general, we use distributions for representation Choosing the best processing is a key issue of knowledge with uncertainty. Information is a tractable representation of context specific For binary information, we may just supply the knowledge. probability. Monitoring is the sub-path from registration to information For continuous information, we supply for instance a normal distribution with a mean and a variance. A small variance implies that we are rather certain During this course we will follow the path from data to about the value. decision: For consistent processing of data, a model is needed. Slide 15 Slide 16 Department of Veterinary and Animal Sciences Department of Veterinary and Animal Sciences Uncertainty: Pregnancy status of a cow Models for monitoring under uncertainty For the replacement decision we want to know the Assessing the distribution of a key figure from the pregnancy status of a cow, but it is (most often) distribution of data (later this lecture): unobservable, so we have only indirect observations: • “Black box” approach • If a cow has not been inseminated, the probability of pregnancy is zero. Statistical quality control models based on time • If it has been inseminated, the probability is 0.4, series analysis: because the conception rate of the herd is 0.4. • More or less “black box” approach • If, after 5 weeks, the cow has not shown heat, the probability increases. Assuming a heat detection rate of Dynamic Linear Models based on Bayesian updated 0.5, the probability increases to 0.7. time series with Kalman filtering: • A positive pregnancy diagnosis will further increase the probability, but only a calving will increase it to 1. • Structured model We need a method for consistently combining the • Data of same kind indirect observations into an updated probability of Bayesian networks pregnancy. • Highly structured models We may use a Bayesian Network model for that! • Data from different sources Slide 17 Slide 18 3

9/3/2018 Department of Veterinary and Animal Sciences Department of Veterinary and Animal Sciences Focus on the monitoring process Basic production monitoring Traditional approach to production monitoring: • Registrations are collected systematically in the herd • Data is entered into a Management Information System (MIS). • At quarterly (or monthly) intervals, the MIS calculates a bunch of key figures which are presented to the farmer in tabular form in report. • The farmer looks at the key figure and decides whether or not to make adjustments to production. The information provided is the list of key figures. We shall briefly discuss how to interpret such key figures in a sound way. Slide 19 Slide 20 Department of Veterinary and Animal Sciences Department of Veterinary and Animal Sciences Data recording Principles of production monitoring Data recording Events: Database • Mating • Farrowing Data processing • Weaning Report with key figures • … Analysis • Statistical • Utility Decision Slide 21 Slide 22 Department of Veterinary and Animal Sciences Department of Veterinary and Animal Sciences A record (or ”registration”) A record, example Event (mating/farrowing/weaning …) Event: Farrowing Identification (sow #/section #/batch #) Identification: Sow # 1234 Registration level (animal/section/batch/herd) Registration level: Animal Time (date or date/hour/minute) Time: February 23rd Property (what is measured) Property: Live born piglets Value (numerical or categorical) Value: 12 Slide 23 Slide 24 4