Yield Prediction Challenge

 
 
Design Challenge
SDP aims to increase Fresh Fruit Bunch (FFB) yield whilst simultaneously reduce FFB production cost via amongst others, deployment of precision agriculture.
To achieve this, SDP aims to:
  1. Develop yield prediction models with the highest accuracy. Simulation of FFB production vs. stimuli / external factors / estate practices. (Predictive analysis)
  2. Identify key factors impacting yield and possible course of actions / recommendations to achieve the best revenue and profit.(Prescriptive analysis)
SDP intends to achieve predictive and prescriptive results on a field-by-field basis in our estates, starting with Peninsular Malaysia.
For this challenge, we will provide historical field data for 5 locations (the size of an oil palm field is 50 – 100 Ha). The expected outputs must include:
  1. 12-month yield forecast (with the breakdown of 1-month, 3-month, 6-month forecasts) with a minimum accuracy of 70% on a yield per field basis for each of the 5 locations.
  2. Output data must be interpretable & actionable by SDP for post-processing (root cause analysis, actionable recommendations etc.).
  3. Provided 120 months of historical data from 2008 – 2017 (Jan – Dec)
  4. Types of data provided (~35 data variables) agronomic factors (e.g. planting material, terrain info) and non-agronomic factors (e.g. fertilizer application, harvesting method, etc)
The final shortlisted participants of this SDP Digital Innovation Challenge will have the opportunity to pilot, evaluate and replicate the deployment of the model within the company.

 Who can participate?
  • Sime Darby Plantation employees
  • Registered Companies
  • University Students / Lecturers


Timeline

Proof-of-Concept  
1. Requirement
  • All shortlisted participants will need to attend a pre-POC briefing session at the beginning of the POC stage.  The date of the pre-POC briefing will be communicated to the shortlisted participants at a later stage.
  • Participants need to declare all technology tools used during each stage of the POC.
2. Data points to be provided
Below are the data points that will be provided by SDP for the model (non-exhaustive):
  • Fertilizer
  • Tree age
  • Pruning cost
  • Weather
  • Foliar analysis parameters
  • Manpower
  • Harvesting method
  • Historical yield
  • Satellite images
  • Planting material
Key Success Criteria
The winning solution will be chosen based on the following items :
  1. Highest accuracy is given by the prediction model (based on R square) and will be evaluated by yield per field.
  2. Prediction consistency.
  3. The cost to develop, replicate and maintain the yield prediction model over a period of 1, 3 and 5 years.
  4.  Ease of adoption by SDP and integration into SDP’s existing infrastructure and practices.
  5.  Clear links to revenue/cost impacts and easy-to-understand output data for users.
Review Process
  • SDP reserves the sole and absolute right and discretion to select the shortlisted participants.
  • Shortlisted participants will be notified of the venue for the pre-POC briefing session before the POC model development stage begins.
  • After the POC model development stage ends, participants must send their final prediction model to SDP for validation & evaluation purposes.
  • The evaluation of proposals will be made by SDP and the decision shall be final. The team of judges consists of representatives from SDP. The model will be judged on the key success criteria as explained previously.
 
submission guidelines

submission format 
 
rules of challenge 
Your shopping cart is empty.