dc.contributor.author |
Farhat, Elham Mohamad |
dc.date.accessioned |
2020-03-27T22:52:14Z |
dc.date.available |
2020-03-27T22:52:14Z |
dc.date.issued |
2019 |
dc.date.submitted |
2019 |
dc.identifier.other |
b23524753 |
dc.identifier.uri |
http://hdl.handle.net/10938/21670 |
dc.description |
Thesis. M.S. American University of Beirut. Department of Computer Science, 2019. T:6985. |
dc.description |
Advisor : Dr. Fatima Abu Salem, Professor, Computer Science ; Members of Committee : Dr. Wassim El Hajj, Associate Professor, Computer Science ; Dr. Shady El Bassiouny, Professor, Computer Science. |
dc.description |
Includes bibliographical references (leaves 73-74) |
dc.description.abstract |
Being able to predict demand by the victims on Red Cross and Civil Defense services plays a vital role in Lebanon which has been reeling under the effect of regional struggles. This thesis attempts to explore data from tweets of the Lebanese Red Cross and Civil defense services to gain insight into the demand following car accidents and _re outbreaks. These tweets published by the Red Cross and Civil defense services Twitter accounts correspond to the SOS calls issued by the victims incurred during the years 2015-2016. We collect the data by scraping Twitter and generate from the Tweets a univariate time series representing demand on a daily basis for the named two years. Each tweet includes the time when the response was recorded, the location where the service took place and the type of accident (car accident or fire outbreak). Before attempting the multivariate predictive modeling employed by several machine learning models, it is essential to explore the classical ARIMA and SARIMA models in order to provide a baseline performance against which other machine learning models can aim to beat. The first part of the thesis provides a comprehensive ARIMA and SARIMA study for forecasting the weekly, biweekly and monthly demand for Red Cross and Civil Defense services in Lebanon using data provided by their respective tweets. Forecasting is trained on the three data sets, and delivers a tool that can be employed by the Lebanese Red Cross and Civil Defense Services to be able to predict the expected demand on a weekly, biweekly and monthly basis. Beginning with an exploratory analysis of the data, we process the time series with regards to its seasonality and stationarity leading up to the actual modeling. We conclude with a comparative assessment between the ARIMA and SARIMA models on new data pertaining to the year 2017. The fact that demand cannot be simplified as a simple, univariate time series phenomenon but is rather actually dependent on weather conditions and public holidays or events, makes it compelling |
dc.format.extent |
1 online resource (xi, 74 leaves) : illustrations (some color) |
dc.language.iso |
eng |
dc.subject.classification |
T:006985 |
dc.subject.lcsh |
Forecasting -- Data processing. |
dc.subject.lcsh |
Red Cross and Red Crescent -- Lebanon. |
dc.subject.lcsh |
Civil defense -- Lebanon. |
dc.subject.lcsh |
Machine learning. |
dc.subject.lcsh |
Time-series analysis -- Data processing. |
dc.title |
Data science approach for forecasting the demand on Lebanese Red Cross and civil defense in case of accidents. |
dc.type |
Thesis |
dc.contributor.department |
Department of Computer Science |
dc.contributor.faculty |
Faculty of Arts and Sciences |
dc.contributor.institution |
American University of Beirut |