rkalm / ann-mallards

Using ANN to predict the movement of mallards

Using ANN to predict the movement of mallards

Names: Alm Robert, Imeri Lavdim

Date: 2022.01.29

Faculty of Natural Sciences

Link: https://researchportal.hkr.se/en/studentTheses/using-ann-to-predict-the-movement-of-mallards

What

This project aims to calculate the movements of mallards, computationally, by using an ANN algorithm [1], We aim to investigate the possibility of increasing the precision of the prediction by understanding the element of choice of living creatures which is affected by a variety of factors that we try here to identify.

Mallards carry the scientific name “Anas platyrhynchos” [2] and they considered as wild ducks that habitat large areas around earth, mostly on the northern hemisphere.

Mallards are living creatures that make choices and decide how they will move, and what are they going to do next, in order to achieve the necessary elements for their survival, (they need to fulfill their needs, like food or sleep, like all the other living creatures do).

The scientific society is expressing great interest on tracking down the movement of living creatures, (for example birds like mallards), as the data can be used for biology or environmental researches.

Research Questions

1. How accurately can we predict the movements of mallards using ANN?
2. What factors can increase the accuracy of an ANN on predicting the movements of mallards?

Why

Many kinds of animals worldwide are threatened with extinction by the loss of habitable land, environmental factors, and of course illegal hunting. A common practice for the protection of wildlife is the study of the animals’ behavior, (movement). Currently, researchers are using devices to track the movements of wild animals, a methodology that requires a lot of resources and effort. Having an algorithm that can predict the movement of animals, based on their environmental factors, can decrease the cost of tracking down the movement of animals, while it can increase the efficiency of the tracking. On the area of sustainability and according to the 17 “Sustainable Development Goals” [3], (SDGs), this project can contribute to the support of SDG15, (Life on Land), and to extend towards SDG14 which is about “Life below Water”.

How

To deliver optimal results, our research needs to take advantage of both experimental methodology and literature research. While, to investigate the accuracy of the ANN algorithm when trying to predict the movement of mallards is essential, literature research can be the key to identify the factors that make a mallard to take a choice.

This research is heavily influenced by the research that published by Daniel Einarson with the title “Hierarchical Models of Anticipation” [4], and because the movement

prediction algorithm follows anticipatory characteristics, we will aim to use the research as the theoretical framework in our efforts to investigate the factors that could increase the accuracy of our prediction.

At the same time, the work of Robert Rosen, with the title “Anticipatory Systems” [5] can serve as a guideline on what we mean by “anticipation” and how we can achieve it, Also, the publication from Daniel M. Dubois, “Introduction to Computing Anticipatory” [6] gives another perspective on how the behavior of a living creature can be simulated.

Understandably, when it comes to living creatures, the computational model for their simulation is not the only factor that affects their behavior, but also the needs of the living creature need to be taken into consideration, so for our case, we need the necessary theoretical background to understand better mallards and their behavior.

The book “Invasive Birds” [7] by M Guillemain, Pär Söderquist, J Champagnon and Johan Elmberg, manages to highlight the basic characteristics of mallards’ behavior, while the article from Richard M. Kaminski and Ernest A. Gluesing for the journal “The Journal of Wildlife Management” with the title “Density- and Habitat-Related Recruitment in Mallards” [8], gives an overview of the migration and the social structure of mallards.

Experimental Methodology for Research Question 1

We intend to run a simulation based on data that is collected by tracking the hourly movements of mallards in a specific area and in specific period of time, and to run the simulation by using an ANN algorithm, running the same spatial and temporal parameters, (season and geographical area), as the tracked data, to investigate how accurate the prediction is going to be.

Literature Review for answering the Research Question 2

While investigating the factors that can affect the decisions of a mallards, we are going to do a literature review on the research about the biology or the behavior of mallards and computer science techniques that could be used to identify these factors.

Expected Results

From Research Question 1 we expect an indicator of accuracy of an ANN algorithm that aims to predict the movement of mallards in a specific geographical area and on a specific time. The indicator will be in the form of an evaluation of the ANN model, and it will be expressed on a scale of percent, (between 0% and 100%).

From Research Question 2 we expect to draw useful conclusions about the factors that affect the decisions of mallards and how we will be able to integrate these choices to a prediction.

References

1. Brownlee J. Machine Learning Mastery. [Online].; 2018 ]. Available from: https://machinelearningmastery.com/how-to-develop-convolutional-neural-network-models-for-time-series-forecasting/.
2. Jobling JA. The Helm dictionary of scientific bird names London: Christopher Helm; 2010.
3. United Nations. United Nations' official page about the Sustainable Development Goals. [Online].; 2021 [cited 2021 January 29]. Available from: https://sdgs.un.org/goals.
4. Einarson D. Hierarchical Models of Anticipation. In AIP Conference Proceedings; 2002; Lund. p. 533-542.
5. Rosen R. Anticipatory Systems: Philosophical, Mathematical and Methodological Foundations. 1st ed. Oxford: Pergamon Press; 1985.