The result of social innovation is all around us. There are signs that social innovations are becoming more and more important. We are submerged by a tsunami of Big Data and data driven innovations. Big Data have been put into practice to help solve social problems. However, small emerging countries in this field are always struggling. This can be due to lack of resources (human and technology) or inadequate use of resources available. We present a new ecosystem for social innovation through Astronomy education via Big Data Analytics. We look at novel methods for educating University learners about Big Data in Astronomy through an innovative pedagogical Education method. Participants will learn how they can apply their various skills to contribute to the betterment of society. We believe that through education at all levels, we can make a paradigm shift for the future generation to prepare them to face the data driven deluge that will present itself once big Astronomical projects such as the Square Kilometre Array (SKA) and the Large Synoptic Survey Telescope (LSST) come online.
The goal of the BArIStA1 workshop aim is to foster research techniques and analysis tools to aid the students in their current area of research and broaden their exposure of topics in astronomy and data analytics. We are also aiming to foster the necessary tools and knowledge that have been successful in the private sector to enable the participants to contribute towards the development of astronomy in their home country as well as communicate their research. Participants will be exposed to multidisciplinary research techniques in an active research problem through tutorship from the highly skilled practitioners and academics. This will be achieved through building their scientific capabilities in Astronomy and Machine Learning via hands on tutorials and lectures. We use Astronomy as a springboard towards building skills and competence around Data Analytics and Big Data:
● Introducing the basic concepts of Big Data
● Introduce the basic concepts of Machine learning and Deep Learning
● Practical applications of tools to a working problem
RFI headed by Dr. Nadeem Oozeer & Olorato Mosiane
The aim of this project was to train the student about Radio frequency interference and how to detect them. There are various algorithm for detection, however, the challenge here was to extract data from the noisy corrupted signal.
One Million Songs Database headed by Dr. Pierre-Yves Lablanche
The Million Song project had two main aspects : data exploration/analysis and prediction, each of them offering many challenges. The first aspect of the project consisted in doing data exploration and analysis which appeared to be very challenging considering the size, the diversity of the dataset and the number of missing values. Correlations and patterns were extensively studied using classic but also advanced unsupervised clustering techniques. Due to the time limit focus was given to the strong correlation between the artist familiarity and the song hotness, diving artist in growing v/s fading as a function of their familiarity. The second aspect focused on the possible correlation between songs lyrics and their expected hotness. Analysis were performed using unsupervised methods to cluster them by “closeness” or similarity in meaningful groups followed by a simple feed-forward neural network to predict the hotness. Although done on a subsample of the original dataset, predictions were found to be surprisingly close to the real hotness but analysis must be done on the whole dataset to draw better conclusion on the ability to predict a song popularity solely based on its lyrics.