Overview
Concurrency’s Data Science practice helps clients improve their businesses through insights that lie beyond the reach of human intuition and traditional modes of analysis.
How Data Science Fits In
| |
Business Intelligence
|
|
Internet of Things (IoT)
|
|
Data Science
|
| |
Gather and present data so business leaders can take action
|
|
Vastly expand data-collection capabilities through sensors
|
|
Apply sophisticated modeling and analysis techniques to generate actionable conclusions that cannot be obtained in other ways
|
Data Science Project Framework
Data Science features a slightly different vocabulary than most information technology domains. As a scientific discipline, Data Science involves developing and testing hypotheses. Therefore, experimentation is a critical aspect of any Data Science Project.
|
Define the Problem
|
What is the desired outcome that is currently unachievable without the rigor of data science, which includes statistical modeling, machine learning, and sophisticated data clustering methods? The answer to this question will provide the basis upon which the project roadmap is designed.
|
|
Define Success
|
What does success mean? Data Science technology allows us to back-test the solution against actual historical circumstances to evaluate the efficacy of the solution. Does the back-test suggest an improvement in efficiency? If so, there is a high likelihood that a ROI is attached.
|
|
Understand the Data
|
Data can only be information when it is understood by an intelligence; i.e., a scientist and/or business user. Therefore, data must be intelligible and must accurately quantify what it seeks to represent. Data exploration and visualization techniques allow us to ensure that certain data characteristics don’t defy our intuition.
|
|
Develop Hypotheses
|
A hypothesis is a summary the methodology likely to yield the solution necessary to arrive at the desired outcome (the educated guess). Hypotheses are an imperative in the data science field because almost all projects in this realm function like experiments. It’s obvious but worth stating that Data Science is indeed a science.
|
|
Design Experiments
|
Once the hypothesis is designed through collaboration with the business user, experimental design is largely up to the data scientist. These experiments revolve around the hypothesis and involve parameter tuning, (computational) modeling, and algorithm design.
|
|
Test
|
The product of each experiment is tested to evaluate its efficacy within the system that it is designed to function. Sometimes this process is referred to as the test data set, which is a portion of data from which the product was blinded.
|
|
Evaluate Results vs. Success Criteria
|
Here, we evaluate the product performance of the product if it were to be implemented in a live production environment. Sometimes this process can be carried out through simulation. Otherwise, it may have to be implemented to run in parallel with business processes.
|
|
Design a Solution
|
Normally, this step involves designing the cloud architecture necessary to handle processes like data ingestion, the computational load, and data output.
|
|
Build Solution
|
Simply put, the data science solution and cloud solution architecture are built for production.
|
|
Operationalize Solution in the Business
|
Finally, turn the solution ON.
|
Data Science Project Examples
-
Sales Forecasting / Supply Chain Management: Manufacturing
Most companies are still using sales forecasting models from the 1980s. Machine learning will make your sales forecasts more accurate and faster, catch correlations no human ever could, and prepare a unique model for every product.
-
Predictive Maintenance: Manufacturing
Downtime is one of the biggest expenses in manufacturing, but typical maintenance schedules are somewhat arbitrary and expensive—you’re paying technicians to look at machines in no danger of breaking down even as actual problems get overlooked. Machine learning allows you to combine sensor data with the power of the cloud to catch problems just before they happen, all while spending less on routine maintenance.
-
Predictive Analytics for CRM
Minimize human intuition in the sales process and get better results. Machine learning can predict the probability that a deal will close based on factors like which account executive is responsible, the timing and length of calls and emails, and even the prospect’s level of enthusiasm.
-
Text Document Prioritization
Do you have time to sort out useful information from 100,000 customer reviews? An AI does. Using Bayesian methods and natural language processing, you can interpret and prioritize court documents, customer complaints and sentiment, tax documents, prioritizing and grouping and much more.
-
Fraud Detection: Finance/Insurance
Fraud follows specific patterns, but the costliest cases are subtle rather than obvious. Machine learning will catch anomalous spending behaviors and flag suspicious insurance claims as soon as they happen with minimal human involvement.
-
Customer Segmentation and Targeting: User Data Required
Digital marketing was one of the earliest fields to adopt predictive analytics. AI allows you to go far deeper than typical demographic segmentation. Machine learning creates such mathematically precise models that tens of thousands of customers can each be served a precisely targeted mix of ads.