<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Ben Stanbury on Medium]]></title>
        <description><![CDATA[Stories by Ben Stanbury on Medium]]></description>
        <link>https://medium.com/@ben_stanbury?source=rss-e76bd1f883b4------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*uDmEQEYvmH6iNJ5s_scfug.jpeg</url>
            <title>Stories by Ben Stanbury on Medium</title>
            <link>https://medium.com/@ben_stanbury?source=rss-e76bd1f883b4------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 00:40:13 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ben_stanbury/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Fit for Data Science]]></title>
            <link>https://medium.com/@ben_stanbury/fit-for-data-science-1850ee583e2d?source=rss-e76bd1f883b4------2</link>
            <guid isPermaLink="false">https://medium.com/p/1850ee583e2d</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Ben Stanbury]]></dc:creator>
            <pubDate>Mon, 04 Mar 2019 15:21:41 GMT</pubDate>
            <atom:updated>2019-03-04T16:51:34.478Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ag_tvymoqpxnyUOLtw9Ntg.jpeg" /></figure><p>A career in data can be rewarding, challenging and fun but, as discussed in <a href="https://towardsdatascience.com/from-data-analyst-to-data-scientist-f67a724ea265">my previous post</a>, is no stroll in the park. Goals, plans and talented peers may get us some of the way, but we need to be mentally and physically ready for the challenge too.</p><p>The suggestions here are, largely, “just lifestyle advice”, but I’ll draw out key aspects relevant to our field, and identify why they are important to us.</p><p>Quick disclaimer — I’m bad (read terrible) at a lot of this stuff. On first glance it seems basic, but it is a LOT easier said then done. Let’s make a commitment to ourselves to do better.</p><p>Okay you’ve patiently read the intro. Let’s crack on with five focus areas.</p><h3>Sleep</h3><p>The bag of tricks we offer includes programming languages, data cleaning, munging, cloud computing, predictive modelling, visualisation…(okay you get the point!). Most of them require DEEP thought to master and apply, requiring a LOT of hard work and concentration. If we slide into microsleeps at the desk, we’re not going to achieve the competence we require, nor deliver our end goals effectively.</p><blockquote>6 hours 45 — nightly sleep of an average Britain (The Sleep Council).</blockquote><p>Most of us are likely skimping an hour or two on sleep. Possible factors may include stress, busy jobs or uncomfortable room conditions.</p><p>So what are key benefits we seek from taking our sleep more seriously?</p><ul><li>Effective learning.</li><li>Better memory and concentration.</li><li>Sound judgement.</li><li>Enhanced problem solving/reasoning.</li><li>Less risk exposure to illnesses and depression.</li></ul><p>Wow that’s quite a list, and hugely relevant to being a data scientist.</p><p>Let’s take an example. Imagine we build a model to predict toxic language in text. We’ll require an appropriate model choice. We’ll want to think imaginatively about new features. We’ll need to optimise for speed. Are we underfitting/overfitting? What is the cause? What steps can we take to achieve a better decision boundary? How many epochs should we run? Perhaps we need a refresher on (say) neural networks, including a few variants we’ve not seen before. This is heavily layered and will require every ounce of our concentration, judgement and problem solving, i.e. most of the list above.</p><p>So what pledges we can make to make to ourselves to do better?</p><ul><li>Establish a <strong>regular sleep schedule</strong>. Let’s get the 8 hours everyone bangs on about. A bit of maths can give us our lights out time here.</li><li><strong>Calming bed routine</strong>. Our much loved brightly lit mobile phones should be put to one side. Let’s also avoid stress, excitement and food/alcohol before bed too.</li><li><strong>Check the room/bed setup. </strong>Is the room temperature okay? Are we happy with the mattress?</li></ul><h3>Attention</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EXJNUtx9u6wiHnprCG6yFg.jpeg" /></figure><p>Checking social media 100 times a day? I’m afraid we need to wind that right in. Our phones can be informative, interesting and fun, but we’re working against a deluge of distractions and our attention is all over the shop. Most of it is unworthy of our time.</p><blockquote>64% say personal smartphone usage at work distracts them (Deloitte’s Mobile Consumer Survey UK).</blockquote><p>In our field success follows discipline and focus. Lets say we’re debugging a complex coding routine, perhaps it involves a loop, creating lists and dictionaries and eventually a data frame. This requires extended concentration. We can’t afford interruptions every few minutes.</p><p>It’s important we understand the back story to mobile phone use too. In the <a href="https://en.wikipedia.org/wiki/Attention_economy">attention economy</a> reward is given to apps, games and services that can capture our mental engagement. This is monetised through advertising and/or sale of data. Our evolutionary reward systems can be exploited by phone and app designers. Company profit is likely to be a bigger driver than our best interests here and we need to take back some control.</p><p>In <a href="https://en.wikipedia.org/wiki/Operant_conditioning_chamber">Skinner’s box</a> the environment of rats was adjusted to manipulate their behaviour. A lever was used to dispatch food. The rat engages in a behavioural response with the expectation of a reward. This reward reinforces the behaviour, and in turn increases the behaviour being observed. The principals here are utilised in social media apps and computer games. It’s worth a few minutes reading up on this.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FNUMa0QkPzns%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DNUMa0QkPzns&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FNUMa0QkPzns%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/afafd8dccceb8fda3fe04837d14bf7e7/href">https://medium.com/media/afafd8dccceb8fda3fe04837d14bf7e7/href</a></iframe><p>Now before we get too down on phones, they do have benefits too. We can call friends, read interesting and educational blogs, listen to podcasts or watch entertainment.</p><p>So what simple steps can we take to strike a balance:</p><ul><li>Turn off noisy notifications and dig through the settings to block low value notifications. Right now I have a Facebook friend suggestion for someone unknown to me, and a news story about Pokemon. These add little to my life. There will also be a “Do Not Disturb” option worth using too.</li><li>Take time away from our phone. Leave it in another room. Go out for a walk without it. Perhaps even schedule some time for catching up with messages and ignore it at other times. Take back control.</li><li>Put the phone aside for the final half an hour before bed.</li></ul><p>Getting a bit less phone specific, what else can we do to improve our concentration? Perhaps we could try:</p><ul><li>Working in timed bursts of 25 minutes. This is called the <a href="https://en.wikipedia.org/wiki/Pomodoro_Technique">Pomodoro Technique</a>. I’ve not tried it myself yet but I heard some swear by it.</li><li>When working in a noisy environment or office we could try using headphones or finding a quiet space.</li></ul><h3>Mindset</h3><p>When you first wake in the morning, it’s a reasonable bet the thought of work won’t fill you with optimism and excitement.</p><p>Work, like life generally, will have highs and lows. We aren’t designed to be blissfully happy all the time. However, there are MANY reasons we should be extremely grateful for our situation. We must embrace that. We work on complex problems and hear wonderful ideas from amazing people. The success of open source gives us access to fascinating tools and materials that enable us to rapidly upskill in areas we wouldn’t have dreamed of making headway in a few years ago. If we can learn to appreciate our good fortune, and get motivated by it then time will fly by.</p><p>What step can we take for a better mindset?</p><ul><li>At the start of the working week be clear on our goals for the days ahead. Focusing on these can be a great motivator and prevent drift.</li><li>At the very start of each day, think of a couple of things we are looking forward to getting stuck in to in order to wet our appetite.</li></ul><h3>Frustration</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/246/1*9HieogwqcPlhTVZtshHwXg.jpeg" /></figure><p>We need to block out negative influences and manage irritation, because it’s bound to be a factor sooner or later. It’s hard to cover all eventualities, so I’ll just throw out a couple of examples — perhaps a colleague has decided to allocate lots of time to low value activity, or maybe your great ideas on the application of machine learning are being stifled by process and traditionalist views.</p><p>How should we manage these feelings :</p><ul><li>Keep professional. When in disagreement be firm and robust. Keep to the facts, rather than being prickly, aggressive or passive aggressive. Try to keep an open mind and understand the position of the other(s) involved.</li><li>Don’t let annoyances zap our time and energy. If we give them permission to do that, we’re becoming a problem ourselves.</li><li>Be proactive in solving and working through issues. Nobody likes <em>complaining for it’s own sake</em>. We need to be productive with that energy and consider how we <em>work towards solutions</em>.</li><li>We shouldn’t copy poor behaviour. Of course it can be very frustrating when our expectations aren’t met, but one of the worst things we can do is mirror this behaviour “My colleague is lazy. I will be too”. Use people we respect as our role models.</li><li>If we find ourselves particularly wound up, which may be possible occasionally, then consider briefly going to a different area to cool off.</li></ul><h3>Exercise</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*6gXh_7O_R9_oTB-eyC8Pyg.jpeg" /></figure><p>It’s widely accepted exercise reduces risk of various major illnesses, while at the same time boosting self esteem, mood, sleep and energy. This isn’t me saying this. This is the <a href="https://www.nhs.uk/live-well/exercise/exercise-health-benefits/">NHS</a>.</p><p>No doubt people will vary in their enthusiasm for exercise, the most important thing is to hit a baseline level.</p><blockquote>66% and men and 58% of women meet the recommended aerobic guidelines (NHS Health Survey for England).</blockquote><p>In addition to exercise, it’s worth being firmly aware that sitting at a desk all day is really not good for us. Our muscles are tightening all day long and a certain amount of discomfort and injury risk may result. In the data science arena there is a considerable variety of skills to master. There may be aspects we wish to learn for fun, or to enhance our CVs. Realistically this may mean additional working outside hours at times, making it even more important to take health seriously.</p><p>So what are our our actions here.</p><ul><li>Ensure we get at least 150 minutes of moderate activity, or 75 minutes of vigorous activity each week. Perhaps the commute can be modified to include a walk or bike ride.</li><li>Getting a well balanced (ish!) diet without too many treats or alcohol.</li><li>Get up from the desk every half an hour. Do some stretching if possible. Consider a walk/run in the lunch break.</li><li>Consider standing for some calls, and whether some meetings can be run as “stand ups” rather than lengthy sitting sessions.</li><li>Check your desk set up is correct and isn’t creating exposure to risks such as back trouble or RSI.</li></ul><h3>Conclusion</h3><p>Well done for reaching the end. Hopefully with a few points to take away. I’ll return to technical posts in future, but hopefully this detour was worth it. If you found it useful then a few claps would be appreciated. Let me know what you made of it, or if you have any of your own ideas. If you have any success that would be great to hear too. Perhaps you can teach me how to do it!</p><p>Good luck!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1850ee583e2d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From Data Analyst to Data Scientist]]></title>
            <link>https://medium.com/data-science/from-data-analyst-to-data-scientist-f67a724ea265?source=rss-e76bd1f883b4------2</link>
            <guid isPermaLink="false">https://medium.com/p/f67a724ea265</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Ben Stanbury]]></dc:creator>
            <pubDate>Sat, 21 Jul 2018 09:50:57 GMT</pubDate>
            <atom:updated>2018-07-21T11:05:18.441Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/830/1*W-2zhyhe0ufFq58a3aOtKw.png" /><figcaption>Source: <a href="https://github.com/matterport/Mask_RCNN">Matterport mask RCNN</a></figcaption></figure><p>How does someone transition from being a Data Analyst to Data Scientist?</p><ol><li>Go to the <a href="https://uk.linkedin.com/">LinkedIn</a> website. Login.</li><li>Click on “edit my profile”.</li><li>Find the words “data analyst” and replace them with “data scientist”.</li></ol><p>Done! Wow that was easy.</p><p>Unfortunately life isn’t that simple. Here we acknowledge the hard yards to build the skills to unlock insight from more-or-less any data.</p><p>There are loads of great articles on starting out in data science (examples <a href="https://medium.freecodecamp.org/a-path-for-you-to-learn-analytics-and-data-skills-bd48ccde7325">here</a> and <a href="https://medium.freecodecamp.org/aspiring-data-scientist-master-these-fundamentals-be7c54350868">here</a>) but less is said about the transition from data analyst.</p><p>Before starting it’s worth me attempting a hand wavy definition of the two roles.</p><p>A <strong>Data Analyst</strong> collects, processes and applies statistical algorithms to structured data in order to yield benefits and improve decision making.</p><p>A <strong>Data Scientist </strong>has similar goals but also has<strong> </strong>robust skills for dealing with large quantities of unstructured data, potentially processing in near real time. They discover important information and are able to clean, process and run advanced algorithms on the data, which may orginate from a variety of sources. They have strong story telling and visualisation skills.</p><p>I regularly encounter talented analysts, keen to begin their data science journey, frustrated by a lack of opportunity and uncertain where to start. This motivated me to write this post.</p><h3>Why become a data scientist?</h3><p>Erm, loads of reasons actually …</p><p><strong>Impact</strong> — Potential to generate enormous business benefit. An opportunity to be heard at high levels and help shape future direction.</p><p><strong>Mastery — </strong>In a fast evolving field there are endless fascinating problems to be solved with a wide range of approaches. For example, building an image recogniser, or a text classifier to identify toxic comments on social media.</p><p><strong>Relevance</strong> — With some predicting that <a href="http://www.bbc.com/capital/story/20170619-how-long-will-it-take-for-your-job-to-be-automated">AI will take our jobs </a>eventually. A safe way to stay above the line it to create automation, rather than waiting to be automated oneself.</p><p><strong>Pay/Opportunity — </strong>Travel around on a big yacht, drink champagne in Jacuzzis. Perhaps not but pay can be relatively good. Good data scientists are rare and in high demand; Let that be you and life should be good.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/467/1*Q6zigHas9qwUFge8Q324tg.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/1*NTxWzxq1EDAOfhi0TZd9mg.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nOZKWwSZrHTR9jgiD38YVQ.jpeg" /><figcaption>Data science — loads to learn (left) , robots taking our jobs (middle) and good pay but remember that The Wolf a Wall Street ended in tears! (right)</figcaption></figure><h3><strong>Am I already a data scientist?</strong></h3><p>Most analysts will have a good foundation, but it takes years of effort to develop skills to comprehensively apply cutting edge approaches on awkward structures and/or large data sets. That said it shouldn’t take long to start having success, and even at a level exceeding our early expectations.</p><p>So what skills are we working towards? There probably isn’t a correct answer to that question, but a sophisticated data science project may have a complex pipeline with many elements. In the first couple of years of data science development one might expect to at least touch upon a reasonable chunk of the following:</p><p><strong>Data Science languages</strong> — Python/R</p><p><strong>Relational databases </strong>— MySQL, Postgress</p><p><strong>Non-relational databases</strong> — MongoDB</p><p><strong>Machine learning models</strong> — e.g. Regression, Boosted Trees SVM, NNs</p><p><strong>Graph</strong> — Neo4J, GraphX</p><p><strong>Distributed computing</strong> — Hadoop, Spark</p><p><strong>Cloud </strong>— GCP/AWS/Azure</p><p><strong>API</strong> <strong>Interaction</strong> — OAuth, Rest</p><p><strong>Data Visualisation and Webapps — </strong>D3, RShiny</p><p><strong>Specialist fields — </strong>NLP, OCR and Computer Vision</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/938/1*lZdcfBG2wm9fSMs4gRi4fQ.png" /><figcaption>Boosted Trees models are popular in data science competitions</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/496/1*GlUko26YhQFiE38H1hxDGQ.png" /><figcaption>RShiny dashboards can be an effective way to develop an interactive means for others to explore data.</figcaption></figure><p>Acquiring these skills takes a LOT of time (probably longer than your degree course). Even the best in the world still have plenty to learn. However, we can always know more and shouldn’t worry that we have limitations. If we can just get a little bit better each day, we all have the potential to reach a considerable level of skill one day.</p><p>Determination and tenacity will serve us better than our raw intellectual capabilities.</p><h3>Plan of action</h3><p>Before doing anything else we need some foundational skills:</p><ol><li><strong>Start with the right philosophy</strong>. Ten years ago it may have been acceptable to wait weeks to be sent on a data software course. Those days are long gone. There are fantastic materials everywhere. Make learning constant. Practice skills continually.</li><li><strong>Learn a language and develop your maths skills</strong> — Python and/or R is where most start. There are enormous numbers of great free training offerings on sites such as <a href="https://www.coursera.org/">Coursera</a> and <a href="https://www.udemy.com/">Udemy</a>. Many Python users like to use <a href="https://www.anaconda.com/download/">Anaconda</a> and <a href="http://jupyter.org/">Jupyter</a> notebooks. Many R users like <a href="https://www.rstudio.com/">R Studio</a>. On the numeric side, the <a href="https://www.coursera.org/learn/machine-learning">Andrew Ng machine learning course</a> and the <a href="http://cs231n.stanford.edu/">Stanford neural network course</a> are both fantastic and thoroughly enjoyable to study.</li><li><strong>Solve a problem</strong> — Preferably a real problem at your workplace, working alongside business experts and data engineers. This is the best end-to-end form of development.</li><li><strong>Enter </strong><a href="https://www.kaggle.com/"><strong>Kaggle </strong></a><strong>competitions</strong> —<strong> </strong>Kaggle tasks are scoped and cleaned but there is no better way to improve model building skills than to do a challenging problem alongside a few thousand other people. Don’t worry about rank. Start with a playground competition and work up from there. Everyone starts somewhere just give it a try.</li><li><strong>Find out what leaders in the field are saying </strong>— Some like to refer to the “rock stars of data science”. This group make fascinating contributions that are well worth your time. Keep an eye out for the likes of <a href="https://en.wikipedia.org/wiki/Geoffrey_Hinton">Geoffrey Hinton</a>, <a href="http://www.andrewng.org/">Andrew Ng</a>, <a href="http://yann.lecun.com/">Yann LeCun</a>, <a href="https://www.usfca.edu/arts-sciences/faculty/rachel-thomas">Rachel Thomas</a> and <a href="https://su.org/about/faculty/jeremy-howard/">Jeremy Howard</a>.</li><li><strong>Use effective ways of working </strong>— Once the foundations are laid start improving your workflow using version control systems such as <a href="https://github.com/">GitHub</a> for deploying and maintaining code. Consider <a href="https://www.docker.com/">Docker</a> for containerisation.</li><li><strong>Communicate effectively</strong> — We need to be able to sell our work. Executives love a shiny demo so always work towards something you can show at the key presentation.</li></ol><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FNKpuX_yzdYs%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DNKpuX_yzdYs&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FNKpuX_yzdYs%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/218902e2f9d85b82270c16ace20a4eef/href">https://medium.com/media/218902e2f9d85b82270c16ace20a4eef/href</a></iframe><h3>Rachel Thomas on Twitter</h3><p>my new blog post: An Introduction to Deep Learning for Tabular Data https://t.co/OrYS1bUAmk</p><p><strong>Get the right setup</strong></p><p>Even with all the skills in the world, if your organisation doesn’t have the right tools and environments then an uphill struggle awaits. There are likely to always be factors out of our control so we need to think about what we can influence.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/620/1*OFisftMy6GegoqVHw2GMdA.jpeg" /></figure><ol><li><strong>Migrate to the right team</strong>— Simplest plan of all. Most medium to large organisations will have at least one small data science team. Join it.</li><li><strong>Partner with the right people</strong>— If a quick job move isn’t possible then engineer an opportunity to work with the most knowledgeable data scientists you know. For example, spot a problem that could be automated and then buddy up with an expert, rather than commissioning them to do it.</li><li><strong>Champion appropriate tools and environments</strong>—Organisations aren’t always sure whether and how to invest in data science tools. They have a busy programmes of traditional analytical delivery to worry about. Security and audit all takes time to develop so they will only hear strong business cases with clear benefits. Take opportunities to champion appropriate environments, tools and training.</li><li><strong>Develop a clear use case</strong>— Understand your business and how data science can be applied. Link the two with a small proof of concept. Use the success to bolster the case for greater access to tools and environments.</li><li><strong>Work with talented people that know more than you — </strong>Be part of a highly skilled diverse team with skills you don’t possess. Not only will you achieve more, you will pick up some of their knowledge by osmosis.</li></ol><h3><strong>Conclusion</strong></h3><p>There is no time like the present to get started. Start your learning today and get working on a real problem as soon as you can. There will be no turning back. Just keep chipping away and surprise yourself with your achievements. Relish the opportunity.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f67a724ea265" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/from-data-analyst-to-data-scientist-f67a724ea265">From Data Analyst to Data Scientist</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>