Use of data mining methods to explore the phenomenon of Fire-Kites launched from Gaza

created by Etzion Harari and Asher Chazan

As of 2018, balloons and incendiary kites launched from the Gaza Strip towards Israeli territory. These measures have led to hundreds of fires in the Gaza surrounding areas. Since launching incendiary means towards Israel is a large-scale effort (Almost 2,000 documented events), we believe that the weather is a significant consideration in launching the means. This hypothesis is based on the fact that parameters such as wind direction and speed affect the trajectory of the balloons.

Therefore, we chose to use meteorological data on the wind direction documented in the Gaza Strip each day. With this data, we will try to pinpoint the source of the balloon launches from the Gaza Strip itself. In addition, we will use additional data mining tools as well as additional Python packages that we have developed (see here) before, for the research.

Landing points of fire kites in Israeli territory. Data includes coordinates in wgs84 geo dd, date and nickname for each landing location.

source: landing points

Meteorological records in the required date range, from the Israel Meteorological Service (IMS).

source : IMS

The work flow is divided into three main stages. Full documentation of the work process can be found in the Python notebook: fire_kite_analysis.ipynb

Processing information and converting data so that it possible to work with it. We will also merged the data tables, so we can analyze them. Finally, we will present some statistics and basic information about the data.

We will use several tools from the field of data mining to:

• point out the Main Targets of the Fire-Kites launchers in Israel area.
• Identify the Launched Areas of the Fire-Kites from Gaza Strip.

To do this, we will use several codes developed by us:

• Cumulative HeatMap Calculation: Code that allows very fast calculation of heatmap from given points. We will use it to identify points in Israel area that have hit by large number of Fire-Kites. Full documentation can be seen here: cumulative-heatmap-calculation

• Clustering by Silhouette: Code that allows clustering according to the parameter that will return the optimal result (according to the Silhouette Score). We will use it to identify clusters of Fire-Kites crash points in the Israel area. Full documentation can be seen here: Clustering-by-Silhouette

• create Sector-Shape polygons: Code that allows calculation of sector-polygon, based on parameters of the starting point, angle, standard deviation of the angle, and distance. We will use it to calculate the estimated area from which each Fire-Kite was launched. Full documentation can be seen here: create-sector-shape-KML-file

Based on the results we obtained in the previous section, we will propose conclusions about the data studied. We will also export and save the results, so that we can perform exploration on them.

• You can find the outputs here outputs
• All the plot created in this project available here: plots
• The Geographical Data save as Google-MyMaps page here: Fire Kite Map

The code uses the following libraries in Python:

• matplotlib

• simplekml

• shapely

• pandas

• random

• numpy

• scipy

• utm

In addition, the project includes the use of three packages built by us:

• cumulative-heatmap-calculation

• Clustering-by-Silhouette

• create-sector-shape-KML-file

The entire research process can be found in the project's Python notebook: fire_kite_analysis.ipynb.

MIT © Etzion Harari & Asher Chazan