Capstone Projects Showcase the Power of Analytics – Insights from the Great Lakes Business Analytics Program
Introduction
The digital revolution has made data ubiquitous, and organizations across industries are harnessing the power of analytics to turn this data into actionable insights. As per a recent report by MarketsandMarkets, the global data science platform market size is expected to grow from USD 37.9 billion in 2019 to USD 140.9 billion by 2024, at a Compound Annual Growth Rate (CAGR) of 30.0% during the forecast period. This rapid growth is fueling the demand for skilled analytics professionals who can extract value from data.
The Great Lakes business analytics program, offered by the prestigious Great Lakes Institute of Management, has established itself as one of the premier programs for kickstarting a career in this high-growth field. A key highlight of the program is the capstone project, where students get an opportunity to apply their learnings to a real-world analytics problem in collaboration with industry partners.
As a visiting faculty member for the program, I have had the privilege of interacting with scores of energetic and motivated students embarking on their analytics journey. Sitting on the evaluation panel for the capstone projects has given me a firsthand look at the impressive work being done by the students under the guidance of their mentors.
In this article, I will share an overview of how the capstone projects work and why they are so crucial for the learning experience. I will also dive into specific examples of projects from recent batches to give you a flavor of the kinds of problems students are tackling. Whether you are considering the Great Lakes program yourself or are an aspiring analytics professional looking for project ideas, I hope you will find this peek into the capstone projects insightful.
Capstone Projects: Applying Analytics to Real-World Problems
The capstone project is the culmination of the yearlong Great Lakes business analytics program. Students form groups with peers from diverse backgrounds and are paired with a mentor, typically from the industry, to work on an analytics problem over a 4-5 month period.
Through the project, students get to experience the end-to-end process of solving a real analytics problem, from collecting and munging data to building and implementing models. This hands-on experience is critical for transforming the theoretical concepts learned in the program into practical skills.
Sriram Alagarsamy, an alumnus of the program, reflects:
"During our course at Great Lakes, we learnt about various techniques like ways of cleaning-up data, multiple algorithms for analysis and modeling, goodness of fit to check model fitness and had numerous sessions on business intelligence. But it is the Capstone Project that introduced us to the real world challenges of applying our learning, testing adequacy, usefulness and subjectivity of certain algorithms. Without Capstone Project our learning would have remained incomplete."
Around 80% of the projects are sponsored by and done in collaboration with industry partners like Value Labs, Whiskkers Marketing Pvt. Ltd., BRIDGE i2i, and Analytics Vidhya. This gives students exposure to the kinds of problems businesses are grappling with and how analytics can help provide solutions.
Shanmugam VaidyanathanVice, President – AI Success & Consulting at Fractal Analytics, one of the industry partners, highlights the value of such collaborations:
"The Great Lakes capstone projects provide a great opportunity for budding analytics talent to work on real-world problems faced by our clients. We have been consistently impressed by the creative solutions and fresh perspectives brought in by the students under the guidance of their seasoned mentors."
A Look at Some Standout Projects
Let‘s now take a closer look at some of the capstone projects completed by Great Lakes students in recent years. These examples, shared with permission from the sponsoring organizations, showcase the breadth of problems that analytics can be applied to and the advanced techniques being leveraged by the students.
1. Predicting Default for Rural 2-Wheeler Loans
In this project sponsored by a financial services firm, a group of students – Kanthimathi Gayatri Sukumar, S.R.Balaji, S.Ramnath, Rudragouda Patil, and Mithur Niranjan – built models to predict credit default for two-wheeler loans in rural India.
Working with a dataset of over 1 million past loans, they engineered 50+ features capturing financial health, income, and repayment behaviors. Advanced techniques like Recursive Feature Elimination and Principal Component Analysis were used to arrive at an optimal feature set.
The group then trained and evaluated a variety of cutting-edge ML models including:
- Logistic Regression
- Decision Trees & Ensembles (Random Forests, Gradient Boosting)
- Artificial Neural Networks
- Support Vector Machines
Here is a comparison of the performance achieved by the different models:
| Model | Training Accuracy | Testing Accuracy |
|---|---|---|
| Logistic Regression | 95.2% | 94.7% |
| Decision Tree (CART) | 99.1% | 93.5% |
| Random Forest | 99.8% | 95.6% |
| Gradient Boosting | 95.6% | 95.3% |
| Neural Network | 95.0% | 94.9% |
| Support Vector Machine | 90.1% | 89.7% |
Logistic regression was ultimately selected for deployment based on its ideal balance of high performance and model interpretability. The model is now being used to optimize lending decisions for two-wheelers, which are a crucial means of livelihood in rural areas.
2. Improving School Outcomes with Prescriptive Analytics
Students Balamuril S, RamKumar R, Senthil J, Sreeraman K and Sriram A took on the challenge of improving outcomes for government-aided schools across 5 districts. By analyzing school performance data and socio-economic indicators, they uncovered insights around the key factors driving student achievement.
The group built a clustering model to segment schools based on their performance profiles. They then formulated a prescriptive analytics model to determine the most cost-effective interventions to drive improvements in each segment. The mixed-integer programming model could answer questions like:
- What is the optimal allocation of a given budget across different initiatives (teacher training, digital learning, infrastructure upgrades etc.)?
- How many schools from each segment should be targeted to maximize overall improvement?
- Which schools should be prioritized for each type of intervention?
Simulations using the model showed that optimizing the intervention mix can drive 2x the impact compared to current ad-hoc budgeting policies, without any increase in spending.
Insights from the project are now being used to inform district-level education policies, with the potential for wider adoption by schools across the country. This is a great example of advanced analytics driving social impact at scale.
3. Sentiment Analysis for Real-time Campaign Optimization
Anjana Agrawal, an IT consultant by profession, took on a rather unique capstone project – helping a political party understand voter sentiment in real-time and optimize its campaign. Elections in recent years have highlighted the growing importance of social media as a bellwether for public opinion.
Anjana collected Twitter data for the party and key rivals using the Twitter Developer API. State-of-the-art NLP techniques like BERT and FastText were used to analyze sentiment towards the key entities. Topic modeling with LDA was used to discover the main themes resonating with voters.
The analysis was refreshed daily to provide a real-time barometer of voter perception. This enabled the party to dynamically finetune its messaging and address emerging voter concerns proactively. With social media now playing an integral role in elections worldwide, such real-time social media mining techniques are increasingly being leveraged by parties across the political spectrum.
4. Generating High-potential Insurance Leads from Web Data
A group comprising Harpreet Kaur, Aneet Sachdeva, and Peeyush Tiwari deployed web mining and NLP techniques to generate high-potential leads for health insurance products.
They first built a web crawler to collect relevant data from online forums, social media, and competitors‘ websites. Advanced text analytics techniques like named entity recognition, dependency parsing, and co-reference resolution were then used to extract structured insights around user demographics, life events, and product mentions.
Unsupervised learning with K-means and agglomerative clustering was applied to this enriched data to discover user segments based on their profiles and content interactions. Segments exhibiting high purchase intent and brand affinity were identified as ‘hot leads‘ for targeted sales outreach.
The model is expected to drive a 3-5x improvement in lead quality and conversions compared to current broad-based marketing campaigns. More importantly, it equips the insurer to be more proactive and contextual in its customer engagement.
5. Optimizing Retail Marketing with Recommendation Systems
Soumya Tiwari, Remina Surendrababu, Arnab Majumdar, Vijayalekshmi L and Siju Joseph partnered with a leading consumer durables retail chain to optimize its marketing interventions.
They analyzed 3 years‘ worth of transaction data containing over 500,000 unique purchases across 20 stores. Graph modeling was used to map relationships between customers, products, and brands based on purchase patterns. Deep learning based approaches like Restricted Boltzmann Machines (RBM) and Autoencoders were experimented with the to uncover ‘latent‘ representations capturing underlying purchase behavior.
The final hybrid recommendation engine combined collaborative filtering on the graph data with deep-learned product representations. It powers highly personalized product recommendations and offer bundles based on each customer‘s purchase history, product preferences, and brand affinities.
By deploying the algorithm across its app, website, and in-store channels, the retailer is driving a significant boost to cross-sell and upsell revenue. Targeted product bundles have driven a 26% increase in average order value, while customer retention rates have gone up by 12% thanks to highly relevant recommendations.
The Increasing Accessibility of Analytics
The capstone projects showcase how the students are successfully applying advanced analytics and AI/ML techniques to solve real-world problems at scale. This is reflective of two larger industry trends:
1. Democratization of analytics through platforms and tools
The proliferation of open-source libraries (NumPy, SciPy, Pandas, scikit-learn, TensorFlow, PyTorch) and cloud-based platforms (AWS, GCP, Azure) has made it easier than ever to get started with analytics. Aspiring professionals can learn and experiment with the latest algorithms without prohibitive infrastructure investments.
2. Shift from ‘models‘ to ‘products‘
As data science matures, the focus of analytics initiatives is shifting from building one-off models to developing production-grade data products that are embedded into business processes. The most impactful work lies at the intersection of data, algorithms, and software engineering.
With ‘analytics for all‘ fast becoming a reality, the onus is now on professionals to build the right skills and experience to tap into the immense opportunities it presents. Programs like Great Lakes and the hands-on exposure from capstone projects are helping create a pool of industry-ready talent to fuel this transformation.
Conclusion
As the capstone projects demonstrate, the Great Lakes business analytics program is producing some stellar analytics talent. Students are graduating with a solid grasp of both the technical skills and the business contexts in which they can be applied to create real value.
For aspiring analytics professionals, the projects offer inspiration for building your own portfolio of hands-on work. Pick a real-world problem that resonates with you, get your hands dirty with data, and apply the concepts you have learned. While tools and algorithms will evolve, this ability to connect the dots between data and business impact will hold you in good stead.
And if you are considering the Great Lakes program, the capstone project is undoubtedly one of the biggest value-adds. The opportunity to collaborate with seasoned data science professionals on an actual industry problem is incredible preparation for a successful analytics career.
Industry leaders also see the capstone projects as a great avenue to tap into fresh talent and ideas. Bhasker Gupta, CEO of Analytics Vidhya and a long-time mentor for the program, remarks:
"At Analytics Vidhya, we have been partnering with Great Lakes for the capstone projects for several years now. It is always exciting to see the students apply their learnings to the problems we are looking to solve and come up with creative solutions by the end of the program. Several of these solutions have found their way into our production systems and are generating real business value for us."
With data becoming the new business currency, analytics skills will only become more valuable in the coming years. A McKinsey report estimates that demand for analytics professionals will outpace supply by 250,000 in the next 5 years in India alone. In this scenario, rigorous analytics programs like Great Lakes will play a critical role in building the talent pipeline to support the next generation of data-driven businesses.
By putting students in the driver‘s seat for solving real analytics challenges, the capstone projects are the perfect bridge to industry roles. For both the students and the industry partners, they are fueling a vibrant ecosystem of analytics innovation with immense potential for impact. It will be exciting to see the ideas seeded in these projects take flight and transform businesses and lives in the years to come.
