Time-Series Forecasting for Humanitarian Aid - VAL2G dataset

We present in this repository the most important result of the project carried out in collaboration with the UNSSC: a new dataset for time-series forecasting on migratory flows.

To the best of our knowledge, no similar dataset currently exists that aggregates the information we report, or that could be used for a similar purpose.

Dataset Description

The dataset created starts from a preliminary assumption: in general, the first countries of arrival in Europe are Italy, Spain, and Greece. This assumption derives not only from the evident geographical conformations of the Mediterranean and the migratory routes, but also from the scarcity of data regarding the rest of the European nations.

Given this assumption, we show in this graph the countries of origin of the potential migratory flows, which have a more consistent impact on the data.

You can directly download the final version of VAL2G dataset from Figshare link.

Features Description

All the data are monthly based: this is the finest granularity that we found online and in addition, for the context in which this project is placed, it would make no sense to guarantee a daily prediction as it would not be usable. The final dataset is composed of the following features:

Feature	Type	Description	Source ID
Monthly inflow	Integer	Migrant from a departure country to the arrival country.	A
Fatalities	Integer	Number of death in the country of departure.	C
HDI	Float	Human development index (i.e. statistic composite index of life expectancy, education, and per capita income indicators).	B
Distance Departure Destination	Float	Distance between capitals of departure and destination country in Km.	D
Percentage of currency change	Float	Currency change rate with respect to the previous month.	E
Sum Inflow	Integer	Total migrants' inflow per country of arrival.	A
Date	String	Month and year.	Key
Destination country	String	Destination country's ISO code.	A
Departure country	String	Departure country's ISO code.	A

The dataset was built merging information from different resources, which include:

ID	Type	Source	🔗
A	Migration Inflows	International Organization for Migration	Link
B	Social Indices	UN Development Reports	Link
C	Humanitarian Crisis	The ACLED	Link
D	Geographical References	Google Maps	Link
E	Currency Change	Banca d'Italia	Link

Note: the IDs are not refered to the ones reported on Figshare.

---

Cloning the repo

To cloning the repo throgh HTTPS or SSH, you must have installed Git on your operating system.
Then you can open a new terminal and type the following command (this is the cloning throgh HTTPS):

    git clone https://github.com/PoliTO-ADSP-United-Nations-Project/humanitarian_aid_dataset

If you don't have installet Git, you can simply download the repository by pressing "Download ZIP".

---

Environment

Once the repo is cloned, some python libraries are required to properly set up your (virtual) environment. They can be installed via pip:

    pip install -r requirements.txt

or via conda:

    conda create --name <env_name> --file requirements.txt

---

Execution

The main.py is the entry point of the execution.
You can run the program in this way:

python main.py

The program was built with a python version >= 3.8: any lower version will not guarantee the correct execution of the software.

---

Please, refere to the official paper for further information.

References

If you find this work useful, pleace cite the paper above as:

@article{TSF_4_ha,
  title={Time series forecasting for humanitarian aid},
  author={Bergadano Lorenzo, Frigiola Arcangelo, Mantegna Giovanni, Scriotino Giovanni, Zingarelli Valerio},
  year={2023}
}