Work package 2

Complex Data Analytics

M6-M48

Objectives

Provide a set of analytics methods to extract knowledge from collected and stored IoRT multi-modal and multi-scale data. Descriptive, predictive and prescriptive analytics methods will be defined according to the end-users goals.

Task 2.1 Descriptive analysis

New methods will be developed for discovering hidden and unknown features and patterns in complex cleaned (Task 1.2) data, by means of AI-based methods for computation of agronomic indexes on plants and trees, and factors of a farming system using data at different spatio-temporal scales. Results will be stored in database systems (Task 1.3).

Task 2.2 Predictive analysis

Methods to predict spatio-temporal data patterns and correlations characterizing plant behavior, sustainable practices, and IoRT device behavior will be developed. (Un)supervised multi-modal and spatio-temporal multiscale AI methods will be proposed on integrated data from different sources (e.g., soil and weather sensors, movement sensors, cameras) (Tasks 1.2-1.3). These methods will integrate end-users' knowledge at the training step (ex: active learning).

Task 2.3 Prescriptive analysis

Methods to help end-users make actions based on historical data and indicators (Task 1.2, 1.3) such as What-If, assessment, counterfactual reasoning methods and simulation models will be developed to explore alternative and hypothetical scenarios.

Task 2.4 Analytics architectures

Different cloud architectures for data analysis will be deployed using data standards to be coupled with solutions of Task 1.3.

Task 2.5 Experiments

Proposals will be put to the test against real data and with real end-users.