• This field is for validation purposes and should be left unchanged.
analytics program enablement for growth-mindset organizations

Posts Tagged ‘data’

The Argument for a Data Science Degree

Data science degreeWith the rise in popularity of “data science” as a term and a concept, journalists are increasingly pointing out the lack of a formal data science degree from any university in the country.

This is important because a well-rounded data science degree could produce graduates who are capable of handling all facets of data analytics and data management, rather than the current custom of companies hiring people who may only be skilled in one area of analytics, and training them for the rest.

Recently, Government Technology took the call for a data science degree a step further and outlined 3 key ingredients that such a degree would have to possess to be effective. Writer Tanya Roscorla suggests:

  1. Multi-discipline – the ideal data science education would include mathematics, statistics, computer science and content discipline. This presents a unique challenge for universities, as in order to achieve this mix in the current popular university structure you might have to cross over between several colleges to achieve the well-rounded curriculum.
  2. Graduate-level – undergrads with a focus on one of the four aforementioned disciplines could pursue a graduate data science degree where they’d learn the other three. According to Roscorla’s interview with Jennifer Lewis Priestley, associate professor of statistics and director of the Center for Statistics and Analytical Services at Kennesaw State University, “Young students can’t study all four things at once at an undergraduate level because they don’t have the absorptive capacity to understand all the concepts.”
  3. Research orientation – big data presents big problems, when it comes to analyzing enormous amounts of data and culling the wisdom within it. A data science degree that teaches research would lead to data scientists who can actively seek solutions for performing analytics on rapidly, massively growing data.

For the full article visit Government Technology.

Are you for or against a formal data science degree?

Ready to take your own data analytics education to the next level? Register for a free data mining webinar with The Modeling Agency, or enroll in a comprehensive data mining and predictive analytics training session.

Are You Giving Your Data the Attention it Deserves?

Data Strategy is an Executive ResponsibilityIncreasingly, companies are seeing a need for executive-level attention paid to their data and the valuable wisdom it contains. But some struggle with understanding who’s “in charge” of data management, and just how important a role data manager should be within the organization.

Smart Data Collective recently tackled these concepts in a piece entitled “Who’s in Charge of Your Data?”. The Smart Data Collective piece cites 8 key needs that all organizations have when it comes to their data:

  • Data governance policy and procedures
  • Determination of data needs
  • Data collection
  • Data management
  • Data security
  • Data flow
  • Analysis of data and modeling
  • Data reporting

And, according to Smart Data Collective, as companies delve into these needs they find that data doesn’t really fit nicely into the realm of IT. Instead, companies looking to make the most of their data may find their best bet is to elevate data management to the executive level where it can impact multiple branches of business operation. The article quotes Thomas C. Redman from the Harvard Business Review, “Different kinds of assets, people, capital, technology, and data demand different kinds of management. You don’t manage people assets the same way you manage capital assets. Nor should you manage data assets in the same way you manage technology assets.”

It also quotes Data Blueprint’s Data Management and Data Administration academic paper, “Management should recognize the DM is continuing to move towards a model in which it is moving away from the inclusion of low-level operations and towards more managerial functions.”

All this is to say that if your aren’t giving your company’s data an important role in your business strategy with executive-level attention, you should be!

Ready to get started with data mining? Join The Modeling Agency for a free hour-long data mining webinar, or delve deeper with one of The Modeling Agency’s predictive analytics training courses.

Mining Your Web Data for Impactful Insight

Data Mining for Web DataIf you have a web site for your business — and especially if you engage in e-commerce — every visit to your site generates important data about your visitors… even if they do not make a purchase. Every action a visitor takes on your site contains behavioral insight such as preferences, tendencies and habits. Studying this data through data mining and predictive analytics methods reveals patterns and trends that you can use to build a predictive model and inform vital decisions about your site and marketing strategies.

 

Properly mining web data is not a simple process, largely because of the sheer volume and lack of organization that plagues most web data. However, with clear objectives and the right recipe for tracking visitors on your site, you can learn tremendous amounts of valuable information from your web logs.

 

Here are the basic steps for successfully planning and implementing a web data mining project.

 

  • Identify your objective. Work with your web development, sales and marketing teams to determine what demographics need to be captured and the right methods to do so.
  • Select and prepare your data. Determine the database you’ll work with and check to see if the data needs any special preparation (for example, normalizing dollar fields by dividing by 1000, converting dates to continuous values, converting yes/no to 1/0 or running log or square transformations on skewed data).
  • Evaluate your data’s structure. This will decide the data mining methods and tools you’ll use. Check the overall structure of the database and condition of the data set, and see if it’s skewed.
  • Format your solution. In what format would you prefer the results to appear (e.g. graph, map, decision tree, etc.)? What is your overall goal for the solution (e.g. to gain insight or to increase sales)? How will you act on the results? The answers to these questions will help you plan the most efficient and effective data mining process while avoiding the need to retrofit, right size or misapply an otherwise productive model.
  • Choose your data mining tool(s). Take into account both the structure and nature of your data and the best method(s) for achieving your desired results or meeting your overall goal. Other considerations are modeler experience, end-user needs, environmental integration and results translation.
  • Design your models. Examine your model error rates and improve them if possible, and see if additional data exists that could help your models’ performance. Decide how many models you’ll need for the job, then test and train your models using a random number seed. Keep in mind that it’s far more effective to apply a good model to a solid strategy, than develop an excellent model that’s strategically misapplied.
  • Validate your findings. Double check your results and submit them to unbiased review to determine if they are correct. Be wary of great model results on your training and testing data by applying the model against a validation set to ensure that the model is generalized – and didn’t memorize the training data. Be prepared to launch a new analysis with adjusted models if the validation results did not stand up.
  • Report your findings. Prepare a report that clearly documents the entire web data mining process you used and the justification for the tools you selected, as well as presents the results and your comments. If you can see clear ways to improve upon the process for future analysis, include your ideas. Be sure to document! You’ll be glad you did later.
  • Incorporate your findings into your business. They may impact best practices, marketing, sales techniques and strategic planning. Prepare to routinely monitor the performance of your models, because all models deteriorate. When your existing models are no longer accurate, you will have to make adjustments or develop new ones. This is a natural part of an ongoing model lifecycle management practice.

 

How often you need to mine your web data will depend on the industry you’re in and factors like how frequently customer attributes change – or what we call ‘data velocity.’ If your industry is very dynamic you may need to refresh your models often. Maintaining fresh models will maintain the edge over your competitors.

 

Ready to learn about data mining in greater detail? Register for an upcoming free data mining webinar, or one of The Modeling Agency’s intensive predictive analytics training sessions.

Predictive Analytics Helps Businesses Get Better at Guessing

Predictive AnalyticsPredictive analytics – the use of data mining, machine learning and other techniques to look for patterns in data and make informed predictions – helps companies get better at guessing customer wants, needs and behaviors. In a recent Computing.co.uk article Jim Manzi of Applied Predictive Technologies explained predictive analytics helps companies get “about three percent better at guessing”. That may not seem like a large figure, and the effectiveness of predictive analytics depends on the size of the data and the scope of the project, but even 3% can certainly translate to a lot of cash in the right situation.

Predictive modeling allows companies in any industry to fine tune their sales efforts based on educated guesses what their customers will respond to best. “What we’re trying to do with predictive analytics is to get a little better at what you’re already doing,” said Manzi. “And so really our technologies could  in theory be used to build experiments and from that build models to say that a customer ID is more likely to respond to the following offer than another customer ID.”

Retail operations, banks, insurance agencies and the entertainment industry are just a few types of businesses that use data mining and predictive analytics to increase sales. For some practical examples of data mining, see this post on The Modeling Agency’s blog.

If you’re ready to learn more about how data mining and predictive analytics can work for your company, register for The Modeling Agency’s next free data mining webinar. Or for a more in-depth, hands-on examination, attend an upcoming data mining and predictive analytics training course.

Make Data Mining Your 2013 Business Resolution

Fireworks for the New YearIt’s time time again – the time of year when everyone makes grand resolutions for the bright, shiny new year. This year, if incorporating data mining and predictive analytics into your business isn’t on your list of resolutions, it should be.

Your business has data, perhaps volumes of data – valuable data that can tell you a great deal about your customers or potential customers, and how to increase sales. But without a way to find the actionable information and insight in that data, it’s useless. That’s where data mining comes into play. Learning data mining – even just the basics – is a goal truly worthy of completion in 2013.

With data mining you can identify patterns and trends in your data and then build and apply predictive models to them, and produce an uncanny prediction for the behaviors that you need to anticipate to create more sales.

On The Modeling Agency blog, data mining has been thoroughly explained, as have the reasons why you need to make it a priority to learn the techniques that will help you unlock the wisdom in your data. Here are some of The Modeling Agency’s posts of note from 2012:

 

If you’re ready to take the next step with data mining in 2013, register for one of The Modeling Agency’s free data mining webinars. Or get to know the topic in-depth with one of The Modeling Agency’s data mining and predictive analytics training courses.

Happy New Year!