Guest Lecture On`Multivariate Techniques In Data Science’

A guest lecture was delivered today, 6th March 2020 at Data Science (M Sc DS-1) Classroom, Central Block, CHRIST, Lavasa by Dr C Muthu, Head of the Department, Department of Data Science, Loyola College, Chennai on `Multivariate Techniques in Data Science’.

At the outset, as part of general orientation, he shared a personal anecdote which mentioned the reason which inspired him to begin a course in Data Science at Loyola College, Chennai. Apart from his personal interest in the field he shared three cases which featured real-life problems with the students. They were reducing wastage in a manufacturing company, planning solar energy usage for smart city and addressing consumer’s grievances by a telecom.

Dr C Muthu elaborated his point of view and said that the optimal and accurate solution should be provided for a problem. He also mentioned that in data science process one ought to spend maximum time on Exploratory Data Analysis (EDA). This deals with understanding the data given and trying to find out the target and outcome variables while checking autocorrelation, multicollinearity etc. in an extensive way. He added that domain study is crucial when a problem needs to be solved, as it helps in listing solutions. It was enlightening to know that generally, 20% of the total time is allocated for modeling purpose in a project. He explained by saying that the algorithm to be used (Machine Learning/Deep Learning/Artificial Neural Network) depends on the context of the problem. It is said that ANN provides accurate results in most cases, but in the above mentioned first two use cases, Random Forest provided the best accuracy. The students were surprised to learn that sometimes simple things can solve complicated problems.

It was quite an informative lecture which clarified few more concepts. For instance, he said that Data Visualization, always, helps one to understand problems and provides solutions in a better way and thus has a better impact. He practically mentioned that solving real-time problems enhances confidence and gives better knowledge. During the session Dr C Muthu listed out multiple roles of a Data Scientist (DS). According to him, a data scientist needs to find ways to tackle all sorts of problems in the project, such as, recording real-time data and taking into account the historical data, what type of data needs to be collected, checking data set and handle missing values and outliers.

He further added that the purpose of his / her job is to get insights from the data and work upon it, such as which variables are more serious in a problem. He articulated that even if there is inadequate data, he /  she is expected to come up with a solution in most cases by choosing an optimal solution for the problem so that profit can be gained using less cost. He also informed that generally an organization has a team of dedicated data scientists who solve a real-time problem. They include people with good statistical, mathematical or computer knowledge background with specific role for someone who has exceptional domain knowledge.

The two-hour long session focused on a discussion of the use cases of Data Science and eliminate fear of pursuing data science as a career among the students. Dr C Muthu discussed topics of linear regression (simple and multiple) and logistic regression using comparison graphs and implementation in regression. He also talked about Reinforcement Learning and ANN including the basic concepts, purpose and process. He concluded his insightful lecture by giving practical knowledge pertaining to handling a real-time project in an industry. His sincere advice to the students was to keep practicing using real-time scenarios even if the results are not visible immediately.