Elder Research的封面图片
Elder Research

Elder Research

IT 服务与咨询

Charlottesville,VA 22,149 位关注者

Data Driven. People Centered.

关于我们

Elder Research is a recognized leader in data science, machine learning, and artificial intelligence consulting. Founded in 1995 by Dr. John Elder, Elder Research has helped government agencies and Fortune Global 500? companies solve real-world problems in diverse industry segments. Our goal is to transform data, domain knowledge, and algorithmic innovations into world-class analytic solutions. When we combine the business domain expertise of our clients with our deep understanding of advanced analytics, we create a team that can extract actionable value from the data. Our areas of expertise include data science, text mining, data visualization, scientific software engineering, and technical teaching. Experience with diverse projects and algorithms, advanced validation techniques, and innovative model combination methods (ensembles) enables Elder Research to maximize project success for a continued return on analytics investment. In 2020 we acquired the Institute for Statistics Education at Statistics.com to provide focused data science, analytics, and statistics training for corporations and individuals. The Institute’s certificates and degrees are certified by the State Council of Higher Education for Virginia, and its courses are approved by the American Council on Education. Elder Research’s Analytics Services are designed to scale based on the unique requirements of each organization and can maximize the client’s return on analytic investment. Elder Research is also a leader in advanced analytic training and offers a variety of training services directed at each of the key stakeholders within an organization. Training builds a common foundation and vision for analytics across business units and lead to the successful adoption, deployment, and maintenance of analytic models within an organization.

网站
https://www.elderresearch.com/
所属行业
IT 服务与咨询
规模
51-200 人
总部
Charlottesville,VA
类型
私人持股
创立
1995
领域
Model construction、text mining、predictive analytics、sentiment analysis、data science、analytics training、outcome-based modeling、fraud detection、cross-selling/up-selling、customer segmentation、anomaly detection、investment modeling、threat detection和training

地点

  • 主要

    701 E Water St

    Suite 103

    US,VA,Charlottesville,22902

    获取路线
  • 2107 Wilson Blvd

    Suite 850

    US,Virginia,Arlington,22201

    获取路线
  • 1362 Mellon Road

    Suite 130

    US,Maryland,Hanover,21076

    获取路线
  • 14 E Peace St.

    Suite 302

    US,NC,Raleigh,27604

    获取路线

Elder Research员工

动态

  • 查看Elder Research的组织主页

    22,149 位关注者

    What happens when bright students and data professionals get together? Lots of great questions, honest conversation, and a few aha moments for everyone involved. ?? It was amazing to recently have another group of University of Virginia students visit our Charlottesville office to learn what it’s like to work in data consulting. Our founder, John Elder, kicked things off with the story of how our company began and grew over time. Then our team shared their own stories—what it’s like to be a data scientist, engineer, technical business analyst, or client engagement manager at Elder Research. ???? The students came with some great questions, and our team really enjoyed the chance to share more about the work we do, how we do it, and why it matters. The day ended with some great food and even more great conversation at dinner. ??? We’re excited to see where these talented students go next. A big thank you to Professors Karen Schmidt and Xin Cynthia Tong for helping to bring this gathering to life!

    • 该图片无替代文字
    • 该图片无替代文字
    • 该图片无替代文字
    • 该图片无替代文字
    • 该图片无替代文字
      +1
  • Elder Research转发了

    查看Robert Robison的档案

    Senior Data Scientist at Elder Research

    Scaling is an underappreciated trick in forecasting. But don’t take my word for it. The winners of the m5 forecasting uncertainty competition said this: “Time-series data often span several orders of magnitude, requiring scaling and normalization to generalize across related series.”[1] Scaling helps a model more easily learn patterns that appear in different series on different scales. It does this by converting different time series to be on similar levels. (This is only relevant for models that are trained on many time series at once) For example, cold medicine sales go up during the winter. By how much? The pattern at larger stores could help a model forecast how much they’ll go up at smaller stores. Let’s say in a large store there’s a 3X increase. This could serve as a baseline for a smaller store that’s more prone to random variation. Normalizing the data to be on the same scale makes identifying these patterns easier for a model. But how should you scale? Thinking and doing a little research, Kimberly Keiter and I came up with a few different ways: 1 – Standardization: subtract the mean, divide by standard deviation. 2 – Min-max normalization: convert all values to between 0 and 1 based on each series min and max. 3 – Divide by the mean: Divide each series by its mean, so that the average value is 1 for each series. 4 – Divide by an aggregated mean per time period: for example, if we’re forecasting daily store-product sales, divide sales by the average sales at that store during that day. Used by the m5 uncertainty winners. In general, scaling reduces variance but increases the bias of a forecasting model. It’s certainly possible to over-scale: compress time series that act differently onto the same values. This is akin to underfitting. I tend to think methods 1 and 2 usually fall into this category. Differencing is worth mentioning here too: predict the difference between values instead of the values themselves. Not really a scaling method, but it’s a similar idea. More common in univariate methods. What methods did we miss? What have you had success with? Let me know! [1] https://lnkd.in/eXuKGxyy

    • 该图片无替代文字
  • 查看Elder Research的组织主页

    22,149 位关注者

    There are lots of jobs out there, but this helps make them great: Getting to know people beyond their job descriptions. That’s one of the things our CEO Gerhard Pilcher mentioned as a core value while chatting with our newest team members. Every time new people join our crew, our CEO, COO Jeff Deal, and founder John Elder love spending time getting to know everyone and sharing the vision for the work we do. ?? Join us in welcoming our new team members, and learn a little more about them! ?????????????? ????????????????, ???????? ??????????????????: enjoys coaching soccer and playing music (currently learning piano) ?????? ????????????, ???????????? ??????????????, ???????????????? & ??????????????????????: loves investing in people’s learning and traveling to new places ?????????????? ????????????????, ???????????? ???????? ????????????????: has been smiling since she joined the team and is a proud mom of her grown daughter ?????? ??????, ???????? ??????????????????: focused on parenting as she celebrates recently having her second child ???????? ????????-??????????????, ???????? ??????????????: loves reading, trying different foods, playing video games, and taking care of her guinea pig ?????????? ????????????, ???????? ??????????????????: was a math teacher and loves all things nerdy ???????????????? ??????????????, ???????? ??????????????????: likes visiting coffee shops, reading, and exploring walking trails What are some things you think make for a good workplace? If you’re interested in joining our team or know someone else who might be, check out our career opportunities: https://lnkd.in/gpPJUkTT #CareerOpportunities #DataScience #DataAnalytics

    • An image with photos of Elder Research's new team members and executive leaders; the image says "Welcome, New Hires!" in a turquoise comment bubble.
  • 查看Elder Research的组织主页

    22,149 位关注者

    We believe AI built with integrity unlocks its full potential. That’s why 100% of our team—technical and nontechnical—completes our internal responsible AI (RAI) course. ???? AI should be unbiased, reliable, and built with accountability in mind. By embedding these principles into everything we do, we ensure AI is not only trustworthy but also a driver of breakthrough innovation. Because at the end of the day, the best solutions are the ones people trust—and actually deliver dependable results. What’s your team doing to keep AI solutions on track? ?? Explore the principles of our RAI framework: elderresearch.com/rai #ResponsibleAI #AIinnovation #TrustworthyAI #AILeadership

    • 该图片无替代文字
  • 查看Elder Research的组织主页

    22,149 位关注者

    March Madness is here, and it’s time to rethink your bracket strategy! ?? Robert Robison says there's a BIG flaw with bracket pools: they incentivize safe picks (no upsets), making things less exciting and strategic. Some have tried to combat this by adding an upset bonus—a reward for correctly picking upsets. ?? But Robert says there’s a challenge with the typical upset bonus: “It’s either not large enough to matter or so large that it’s senseless to pick anything but an upset.” Check out Robert’s blog for insights on how a different approach to the upset bonus can create a fairer scoring system, where both risky and safe picks are valued, making your bracket more exciting and competitive. ???????????? ???????????? ???? ?????? ???????? ????????: https://lnkd.in/eEuNEYzj #MarchMadness

    • An image with a basketball court in the background. In the foreground is a photo of the blog author, Robert Robison, and the title: Breaking the Tyranny of Chalk Brackets in March Madness.
  • Elder Research转发了

    查看Tom Shafer的档案

    Principal Data Scientist at Elder Research

    Elder Research analytics link roundup, March 10–14, 2025: What does “overfitting” mean, anyway? https://lnkd.in/eaArJ-Tc 1. “Use a test set to select among the models that fit your training data well. It’s not that complicated.” Part of a larger series on?Ben Recht’s blog that sparked good discussion. https://lnkd.in/eqkKVHCa 2. Simon Willison’s recent review of large language models in 2024. I’m reading more and more of this recently—and trialing ChatGPT Plus and Cursor. I'm still mostly ambivalent about these things. https://lnkd.in/e8sDb6Bj 3. Check out these wild graphs showing how StackOverflow queries for things like “R” and “Pandas” have been wiped out in the ChatGPT era. I wonder where this all ends up.?Robert Robison?has the right take, I think, which is to wonder where the equilibrium state is: Are we going to get to a place where LLMs totally replaces StackOverflow, etc.? Will we find a spot where LLMs can answer the easy questions, leaving us to talk to actual humans for the harder ones? Or (I fear), will LLMs put StackOverflow out of business and then we won’t have a centralized place for these kinds of discussions anymore? https://lnkd.in/est_jF3p 4. “The State of Machine Learning Competitions, 2024 Edition”: I was surprised at the number of competitions and the total prize sizes. A pretty healthy variety of solution types, too; it isn’t all deep learning all the time. https://lnkd.in/ejv2KdzG

  • 查看Elder Research的组织主页

    22,149 位关注者

    ????%—that’s the percentage of use cases most analytics applications can handle. ?????? ???????? ?????????? ???????? ???????????? ??%? At business scale, those data anomalies can translate to millions or billions of elements that may be tied to other datasets. And that poses a big challenge, says Data Scientist Tom Shafer: “This mixing of regular and irregular data can be a serious problem for machine learning or AI models that only expect to process typical data.” ?????? ????????’?? ?????? ????????????: “Sometimes anomalous data don’t stem from data-entry errors, storage corruption, or faulty manipulations. When anomalies correspond to fraudulent or otherwise unethical actions, worrisome sensor readings, or physically impossible entries, more value might come from identifying and remedying anomalies than from modeling the normal cases.” Want to uncover valuable insights in your data? ???????? ??????’?? ???????????? ???????? for tips on using anomaly detection effectively, and find out what he says is key to matching the right solution to your business challenges: https://lnkd.in/ehuv9JwM #DataScience #DataAnalytics #AnomalyDetection

    • An image with a blue background and a photo of the blog author, Principal Data Scientist Tom Shafer. Below Tom's photo is the blog title: Business Insights Meet Analytics Skills in Anomaly Detection.
  • 查看Elder Research的组织主页

    22,149 位关注者

    Considering a leap from academia into data science? The math can add up to a rewarding career—just ask Scott Atkinson. He’s sharing his journey from pure mathematics to data science in this series. Follow along for insights. It’s never too late to make a shift! ↗?

    查看Scott Atkinson的档案

    PhD Mathematics | Data Scientist | Research Scientist | Machine Learning | Deep Learning | Python | SQL

    Math to Data Part 1: The beginning To give a clear picture of my journey from math to data, I’ll start with some background on what brought me to start considering this shift. I obtained my PhD in mathematics from the University of Virginia in 2016 and was 100% committed to pursuing a career in academia. After graduating I took a postdoc position at Vanderbilt University followed by a second postdoc at the University of California Riverside (two moves, over 2500 total miles). I was thoroughly enjoying my research in operator algebras (von Neumann algebras and C*-algebras), and I loved teaching courses (some favorites include Differential Equations and Game Theory). At the same time, my wife and I were growing our family. Throughout these postdocs I hunted for tenure-track positions during each hiring season. I was in my first year of my postdoc at UCR when COVID-19 changed everything. As I continued applying for tenure-track positions it was evident that the pandemic was further reducing the already limited number of tenure-track openings, which had the effect of significantly shrinking my choices in location and position.?It was a tough pill to swallow that even with strong teaching (winning a couple teaching awards) and solid research (publishing in selective, highly-regarded journals), I still didn’t have much say over where I would end up. At this point, my wife and I were tired of being “at the mercy of the market,” so in Summer 2020 I began to seriously consider alternative careers where we would have more of a say over where we would build our life.?After some research into math-adjacent careers for PhD mathematicians, I was drawn to data science and machine learning because of the applications of mathematics and statistics to deliver a wide variety of real-world solutions. Also, compared to mathematics tenure-track positions, DS/ML positions were absolutely abundant (and hiring occurred year-round).?So while teaching my summer class over Zoom, I made the decision to begin working toward acquiring the skills and knowledge required for a DS/ML career. Choosing the next phase of my career was only the first step in my journey from math to data. In my next post, I will discuss the next step I took: reaching out. For those of you who made a career switch into DS/ML, what factors influenced your decision?

  • Elder Research转发了

    查看Kimberly Keiter的档案

    Senior Data Scientist at Elder Research

    If you're building a time series forecasting model and you'd like to test machine learning algorithms, I encourage you to give MLForecast a try: https://lnkd.in/eWiRMk9D This week, the Commercial team at Elder Research had our monthly Datathon! This is a time when our team gathers to explore interesting data, try out new techniques, or (as in this week's case) try to win a friendly competition by building a model with the best performance. This month's topic was selected by Matt Bezdek and our objective was to forecast crop production for 14 different crops over 6 different countries. I leveraged the MLForecast framework to develop a couple of models, and the capability I find especially handy is how you can specify feature and target transformations directly in the model object. For example, here's how I specified the number of lags I wanted in my model, the lag transformations I wanted to apply, and the target transformation I wanted to apply: model = MLForecast( ???? models={ ???????? 'lgbm': lgb.LGBMRegressor(), ???? }, ???? freq='YS', ???? target_transforms=[Difference([1])], ???? lags=[i for i in range(1, 11)], ???? lag_transforms={ ???????? 1: [ ???????????? RollingMin(window_size=4), ???????????? RollingMean(window_size=4), ???????????? RollingMax(window_size=4), ???????????? RollingStd(window_size=4) ???????????? ] ???? }, ) I encourage you to check it out the next time you need to build a machine learning model for time series forecasting! P.S. Shout out to Paige Spell for winning the competition!

相似主页

查看职位