After a hundred applications, interviews and enrolled students, our database shows us that the greatest interest in Data Science is among economists and those who work actively in the financing sector.
There is nothing strange here if you take into account the amount of data that can be applied in today’s business, and how much is the possibility of optimisation, based on them. Therefore, bankers and financial analysts are in the lead, as a profile of candidates for our Data Science Bootcamp.
Team 4 from the first group of the Bootcamp was composed of this kind of individuals. Analytical, curious, well versed in statistics and mathematics. And now, with a strong technical skill in the Machine Learning field. Ivona, Riste, Done, Jasmina and Marija are behind the latest project “Bank Marketing with Socio-economic Context”. This is a combination of sophisticated financial techniques and advanced Machine Learning.
If you want to work on projects like this one and add it to your portfolio to provide you with great advantages in the labour market, apply to the Data Science Bootcamp.
Brainster: Colleagues, first of all, congratulations on completing the program and successfully finished the final project! Let’s start from the beginning. Given your finance background, how was the transition, from that “world” into the advanced technologies, that came from Data Science?
Before we started the program, we prepared ourselves as we knew that a long and interesting year, full of learning, awaits us.
Of course, the transition in this area is one big challenge for us, to interpret the analytical capabilities that we develop with years through long working experience, into a modern and sophisticated way.
Data Science Bootcamp – Final Project: Bank Marketing | Team 4
Brainster: Your team worked on a project called “Bank Marketing with Socio-economic Context” based on Machine Learning. Explain with more details the motive behind this project and its benefits?
The dataset was based on historical data got through telemarketing from the contacted customers, to conclude agreements for depositing their funds.
The purpose of the project was to create a model that would predict whether a particular client will enter into a deposit agreement or not. And we got this by using multiple algorithms.
On the other hand, if a telemarketing campaign is reactivated, it would target customers who would most likely deposit funds. The campaign itself will take place in periods that are most favourable for a deposit, according to the historical data from the previous campaign.
Brainster: Unlike the other teams in the group, you didn’t have to create your dataset. You worked on a ready-made dataset from a Portuguese bank that contains data of contacted clients from 2008 to 2010. Tell us more about the feature analysis of this data, their refinement, visualisation and other processes.
Feature engineering is a very important part for further data processing and of course for the result. By analyzing the characteristics of the clients and the data on the conducted calls, we had obtained results. What and how they influenced the conclusion of the deposit agreement.
There were no missing data in the dataset, but only incomplete data that was less than 5%. We didn’t delete them but sought their correlation to fill in that data as well. Regarding outliers, we also decided not to remove them because they were real customer features.
We made a lot of visualisations that helped us get a more realistic picture of the given characteristics, their representation and the impact on the target variable.
Data Science Bootcamp – Final Project: Bank Marketing | Team 4
Brainster: You may not be faced with creating a dataset, but I believe you have faced other kinds of difficulties during the work? You mention a problem with unbalanced datasets in the documentation. How did you deal with this?
The distribution of the target variable in the dataset was indeed 88% versus 12% in favour of not signing a deposit agreement. Here we used different techniques to get the best results. In the end, we decided that it was most convenient to make a balanced dataset from an unbalanced, with the help of various techniques that already exist in the field. For this type of problems, many studies and papers helped us to be able to know, in which direction we should move, according to the research that was already done.
Brainster: On the other hand, how did you cope with the pressure and time you faced for making the project? Teamwork brings many benefits, but it also requires a lot of synchronisation and equal commitment of all members.
The teamwork was harmonious. Everyone participated in what is their strong point. We all contributed to the project development, to be with its best quality. This is how professional Data Science teams work.
Brainster: The development of the project came immediately after the Machine Learning and Big Data modules. As professionals working in finance and economics, how difficult is it for economists to master Machine Learning and Data Science in general?
There were moments of admiration and moments of frustration. Some things we mastered more easily, and for others, we needed more time. But the desire to learn and master new skills was always present, and guided us throughout the year, during the Bootcamp.
Brainster: Given the demand for data staff, does the new skillset arouse your curiosity to try out in another sector? Or you are staying in finance and economics? And do you have a specific domain in Data Science that you would focus on? Such as BI, Data Analysis, ML Engineering, etc.?
The spectrum of Data Science domains is wide. But you have to start somewhere and to go deeper, and only work in this area. We plan to start working and improve the acquired skills as Data Analyst, BI specialist and ML, each of us in his field, as well as for the project itself.
Brainster: Most often the narrative how difficult the program is it, comes down. But we already know that. Was there some module that you expected to “exhaust” you, but in the end, it turned out to be much easier than you expected?
All modules have their weight. But if we have to choose, Power BI was the easiest for us. Through visualisation, we know more easily what it is the expected output. In the Machine Learning part, the expected moment of conclusions was reached. At this point, all previous knowledge and learning gained its weight. Here came the “aha” moment. Because all the trouble that we went through, on statistics and databases, now gain its value.
Brainster: Knowing the whole process, from this perspective, what would you change in the approach of studying Data Science? What would you advise the current and future students at the Data Science Bootcamp?
All students who have already enrolled, have our vision and urge to master what they have begun to learn. We will encourage potential students to decide because they have nothing to lose. They can only gain knowledge and skills that are currently, and even more in the incoming years, to be in high demand in the labour market.
If you have ever thought about a future-proof career in Data Science, have a look at our remote Bootcamps.