CHURN PREDICTION IN THE TELECOMUNICATION INDUSTRY
Annual churn rates in the wireless industry are currently about 10%, while customer base grow more slowly at about 5% annually. Reducing customer churn can have a large impact on customer base. In addition, higher costs associated with acquisition of new customers highlight the need of telecom companies to reduce churn rate in order to decrease costs and increase profits. Data analytics can yield important insights about customer behaviour and may contribute to churn reduction.
A data set of 3,333 US wireless customers, including 483 churning customers, was dissected using regression and decision-trees to reveal:
1. High churn rate among customers with international plans.
2. High churn rate among intense day time users
It is likely that higher bills are driving some of these customers to look for cheaper options.
3. Customers with four or more service calls are more likely to leave the company. Companies should improve their service call centers to resolve customer issues in fewer than three calls.
Analysis of efficiencies in university systems
Comparative analysis of efficiencies in tertiary education systems is challenging due to differences in reporting and lack of standardized tests across nations. I used university revenues as a common financial inputs and the number of graduates as outputs to populate a simplified production model for a crude measure of efficiencies in higher education systems. Three independent lines of evidence suggest that Canadian universities are operating inefficiently. First, analysis of nationally aggregated university data to populate a simplistic production model as suggested above indicates poor performance in the Canadian tertiary education system. Canadian higher education is about 10% more expansive relative to peer countries. In four out of four years examined, the Canadian higher education system was operating at decreasing returns to scales, indicating wasteful operations. Second, these results obtained at the macro, national, levels were mostly corroborated using a representative sample of Canadian and Australian universities. It was found that Canadian universities operate at average 76% efficiency, while Australian universities operate at 86% efficiency. Even worse, most Canadian universities sampled were operating at decreasing returns to scale. Third, examining the growth rates in revenues and enrolments in the Canadian education system reveals higher growth rate in revenues relative to enrolment rates.
My recent projects are at the interface of life- sciences, data analytics, and machine learning.
I enjoy the entire process, from deep analysis of the problem, to the critical survey of potential alternative solutions, to the creation of a final solution and its presentation.
Predicitng Soy yields **First prize award in the Syngenta AI challange**
Source-Sink Relationship Management in Ranunculus **Prize awarded by Syngenta**
Geospatial data mining in gold exploration
Approaches for Medicinal Plant Sorting – the Good, the Bad and the Ugly
High-Throughput, High-Purity Cytoplasmic DNA Isolation from Mature Seeds ** Prize Awarded
Novel Antimicrobial Technology Needed For Liquid Antacid ** Prize Awarded
Naturally Green and Healthy - Natural Grub Control for Lawns
The goal of this work is to develop a predictive model for selecting elite soy variants for commercial production. Current breeding practices for new soy variants require rigorous evaluation over three stages of field tests, corresponding to three successive growing seasons. We propose to leverage machine learning methods for identifying high yielding variants using remote sensing and soil features. To support this proposition, we trained an ensemble of fifteen decision tree models, one for each relative maturity band. Collectively, our models identified fifteen elite varieties from 21 predictive variables to forecast soybean yields in 2015 at 58 test locations. This method can boost commercial soy yields by about 5% and shorten the time for commercial variant development.
For this project you can find:
BIG DATA ANALYITCS - PARKING TICKETS IN TORONTO
I analyzed a big dataset of more than two million parking tickets provided by the Open Data initiative in city hall. I used SAS for data mining and Excel pivot tables for visualization. Here is my presentation.
The hottest spot for picking a parking ticket in Toronto is none other than Sunnybrook hospital. In 2015, 9,076 parking tickets, two thousand more than the next hot spot of Yorkdale mall. Ticketing in Sunnybrook is trending up, a 13% increase, in comparison to 2014 records. Ticketing revenue for the city from the Sunnybrook cash cow were $273,150 in 2014 and $241,555 in 2015. Perhaps a small drop in the large bucket of Sunnybrook budget, yet an indication that something is not working for the benefit of patients and visitors in the transportation and parking system at Sunnybrook.
Parking tickets near hospitals is not exclusive to Sunnybrook. Geocoding the street addresses of the top 1,000 ticketing spots reveals interesting insight about ticketing habits. Exactly 6,719 tickets issued in the streets surrounding St. Joseph's Health center in 2015. Cumulative parking ticket revenue from St. Joseph's for the city was $243,795, another tax on patients and visitors.
Controling budgets and sustaining efficiency in US hospitals
In the aftermath of the Affordable Healthcare Act, public expenditures in in US hospitals climb while outputs are stagnant. Dr. Francis S. Collins, the director of the National Institute of Health (NIH), is invited to a congressional committee to offer strategies for budget control and efficiency improvement in the hospital sector. Dr. Collins intends to use financial and operational hospital data collected in Washington State to identify hospitals operating at sub-optimal efficiency and offer improved budget constrains for this hospitals. Facing proposed 20% budget cut in public hospital spending, the objective of Dr. Collins is to defend the budgets of hospitals operating efficiently, curb budgets in hospitals operating wastefully, and boost the budgets of growing hospitals. This is a deep dive into operational data in hospital and econometric analysis of input and output factors using data envelope methodology. See my report here.
Financial Evaluation of Loblaws bid for Shoppers Drug Mart
Team analysis of SDM prior to the Loblaws bid reveals a large discrepancy between the rate of dividends growth and the rate of revenue and earnings growth in the company. Notably, net earnings slightly declined over the last three years, despite a modest increase in revenues. SDM management were able to maintain the stock price and high dividend rate by drastically reducing capital investment and through deployment of an aggressive buyback policy that eliminated almost 8% of company shares in three years. This analysis leads us to doubt the potential economic growth of the company and the sustainability of maintaining a high dividend growth for SDM over the long term in current conditions. In light of this information, the Loblaw evaluation of the SDM stock at $61.54 corresponds with our optimistic scenarios of dividend growth and discount rates. Shoppers Drug Mart shareholders were right to accept this offer in 2013, since there is a considerable risk that the SDM stock was overvalued by the financial markets and could have experienced a downwards correction in subsequent years.
I am an entrepreneurial data scientist and problem solver. I won several awards for my work with life-science companies, including agricultural and pharmaceutical companies. I worked on drug development projects at the Hospital for Sick Children as a post-doc after receiving a PhD in Medical Genetics at the University of Toronto. An MBA at Ryerson University with a focus in the Management of Innovation and Technology compliments my skills with business intelligence, business strategy, and presentation experiences.
Copyright @ All Rights Reserved