OLX Group is a global online marketplace operating in 45 countries and is the largest online classified ads company in India, Brazil, Pakistan, Bulgaria, Poland, Portugal, and Ukraine. It was founded by Alec Oxenford and Fabrice Grinda in 2006.
A platform that connects buyers and sellers in more than 40 countries and has hundreds of millions of customers per month faces many challenges that are to some extent similar but also somewhat different from online retail.
There are main 3 challenges:
Challenge 1: User experience. When the user navigates the platform and what are the recommendations and the results when doing searches, etc.
Challenge 2: Identifying what is that makes some advertisements much more liquid (easy to sell) than others.
Challenge 3: The reminder after purchasing the items. Predicting if an item is sold 15 days after its entry into the system. If you want to learn more of predictive Analysis, register our predictive analysis on Python here.
As part of the solution for good user navigation and browsing experience, it is useful to have a good estimate if a specific advertisement has been already sold so that we don’t show it again in the recommendation or search output. This is a probabilistic time-series prediction problem. Another important aspect connected to the previous case is identifying what is that makes some advertisements much more liquid (easy to sell) than others. For this particular case, understanding how the model is making decisions is really important as the outcome can be provided to the sellers in order to improve the liquidity of their advertisements. For the reminder of we will focus on this specific liquidity prediction problem, predicting if an item is sold 15 days after its entry in the system, and we will use XGboost and eli5 for modeling and explaining the predictions respectively.
XGboost is a well-known library for “boosting”, the process of iteratively adding models in an ensemble of models that target the remaining error (pseudo-residuals). These “week learners” are simple models and are only good at dealing with specific parts of the problem space on their own, but can significantly reduce bias while controlling variance (giving a good model in the process) due to the iterative fitting approach followed in constructing this type of ensemble. The data we have available for this problem include textual data (the title and textual description of the original advertisement, plus any chat interactions of the seller and potential buyers), as well as categorical and numeric data (the category of the advertisement, the brand and model of the item, the price, number buyers/sellers interactions for each day after the entry, etc.). The data sample we are using here is a relatively small part of the data from some countries and categories only, so in many of its properties, it is not representative of the entire item collection. Nevertheless, let’s start with some basic data munging.
The histogram of the day that an item was sold is shown above. We can easily see that most items are sold in the first days after the respective advertisement is placed, but there are still significant sales happening a month later as well. With respect to the day, an advertisement is added to the platform, we can see that there is a peak on weekends, but other days are roughly at the same level. Finally, with respect to the hour, an advertisement is added to the platform, we can see in the figure below that there is a peak around lunchtime and the second peak after work hours. One way to capture more complicated relations is to use the pair plot functionality of the seaborn library. In this case we will get the combinations of scatterplots for the selected columns, while in the primary diagonal we can plot something different, like the respective univariate distributions. We can see that the number of buyers interaction in the first day is a strong predictor if an item will be sold early or late. We can also see that category it is a very important predictor as well, as some categories, in general, tend to be much more liquid than others. Now that we are done with the basic data munging we can proceed to make a model, using the XGboost library.
Using a hyperparameter optimization framework we can find out the hyperparameters that work best for this data. Since we are interested also on the output confidence of the prediction itself (and not only on the class), it is typically a good idea to use a value for min_child_weight that is equal or larger than 10 (given that we don’t lose in predictive performance) as the probabilities will tend to be more calibrated. The ranking of the features from the XGboost model is shown in the figure above. Although feature ranking from tree ensembles can be biased (favoring, for example, continuous or categorical features with many levels over binary or categorical features with few levels) and in addition, if features are highly correlated the effect can be split between them in a non-uniform way, this is already a good indication for many purposes. Now we select one specific instance at prediction time. Using eli5 we get an explanation of how this instance was handled internally by the model, together with the most features that were the most important positive and negative influences for this specific sample.
As we can see the sample was classified as being liquid, but still, there was some pull down from the text properties (title length, title words, etc) which we can use to provide guidance to the seller for improving the advertisement.
Check out the other predictive analysis article here and register now at Beyond Machine: Machine Learning Predictive Analysis on the 27th of March.
We have conducted some research online that we found out the major obstacles and challenges to work as data scientist or to switch a career to be the data scientist?
#Don’t know where to start or where to learn?
#No Network to find a job or company you like?
#No past experience working as the data scientist? Even you know how to write the algorithm?
Looking for new challenging jobs? Just enjoy algorithm or mathematics? Want to start your own startup? Have a special problem, want to look for the solution? Or just to have a better job with more money?
It is like running a startup, it takes a lot of energy, time, dedication, motivation, discipline, sometimes even amount of money. You can learn all the best deep learning courses from those famous professors and experts all over the world, however, when it comes to soft skill, you need a mentor to help you, if he/she has even extensive networks, that is even better.
Andrew Ng is one of the best instructor and teacher I have ever heard and seen, he is perfect articulated, patient, and full of passion to teach as well. I personally took a couple of his classes online at Standford. (the course was really simple enough for anyone who is determined to start!
Btw, I sucked at Linear Algebra during high school 🙁
If you like more entertaining ones, the Youtube star: Siraj Raval also has Youtube channel and Udemy courses online to follow and learn, his goal is to use machine learning to build anything you can be dreamed of. After enough time for studying and online learning, it is always great to apply to the topic you are especially interested.
Don’t forget the traditional way, reading a book: The Bible of the deep learning which was written by Ian Goodfellow and Yoshua Bengio and Aaron Courville.
Certainly, if you like more personal touch, we offer Deep Learning Bootcampin January in Berlin with the topic of NLP, time series forecasting, and computer vision. Our instructors are from IBM and the famous research institute Max Planck Institute. Romeo Kienzler, he also has Youtube channel and is one of the most successful Deep Learning instructor in Europe.
This is one of the most common questions for beginners. We can have one more article to go deeper into that in the future. For now, choose a language that you are the most familiar with or the simplest for you. If you are completely new, then choose in between the data handling capabilities, advancements in tool, career chance, deep learning support. From my personal experience, I was consulting a machine learning startup with one Python scientist, one R scientist, every time when I had problem with data, I asked Python scientist to write me a quick code to fix my problem while R could not really deal in a such an easy way. Python becomes love of my life for now.
Warning: do not just take deep learning and NLP courses, but forget math is the real foundation. Take also some linear algebra courses also or at least linear algebra.
Don’t waste too much energy and time on it, especially a lot of them come with free alcohols…Check the topic before you attend also attendees, are they your target audiences or not really? Or sign up some recruiting event, there will be tons of recruiters and HR managers from companies. (We are launching our First Recruiting Day focused on the data team recruitment in March 2018)
P.S. M.I.E is currently re-branding to Beyond Machine, our mission to connect and boost AI ecosystem through connecting and training. We will launch series events in 2018, following with Deep Learning Bootcamp, Data Analytics Workshop, Recruitment Day, Self-Driving Workshop, Machine Learning Week. Furthermore, we will launch also outside of Berlin and bring more corporation and partnership exchange in and outside of Germany. If you are interested in any kind of sponsorship, partnership, just simply drop us a line.
WHY SHOULD YOU COME?
2. We are the world FIRST OPEN AIR MACHINE INTELLIGENCE SUMMIT, we had two last evening summits in rooftop bars in Berlin and Paris and this summer will be held in a Biergarten. We know boring conference rooms kill creativity.
3. We can awesome speakers confirmed to talk on the stage.
5. We also offer workshops session hosted by IBM Watson Chief Data Scientist, Romeo Kienzler, he is also instructor at Coursera.
6. We are INTERNATIONAL, we are the first media&community from Berlin expanding outside of Berlin and had speakers from London, Paris, NYC, Amsterdam. Zurich…
More info, please visit our website.
7. Early bird tickets are going to be out soon! Hurry up before the end of May.
Beyond Machine (rebranded from M.I.E) was spawned from Lele and Irene constant frustration during the founding of their AI startups. In the end of 2015, they both left their jobs at rising mobile ad tech and product companies. Lele first started SoCrowd and pivoted to Deckard after 3 months. Irene wanted to tackle the challenges of visual recognition.
It quickly became apparent that there was a need for a more developed community an outlet for media around Machine Intelligence. After running a fruitful and inspiring Evening Summit, they decided to take Beyond Machine. to the next level, founding a media company.focus on the training, re-education, and networking in the field of AI and innovative technology. Irene decided to leave FindEssence, the first company she co-founded and push forward the growth of Beyond Machine.
Beyond Machine’s mission: To connect people in AI industry globally, bringing profound and engaging content and to start a conversation about job substitution issues.
The blurb of M.I.E Summit 2017:
M.I.E. Summit Berlin 2017 is the World’s first open-space machine intelligence summit, which will be held on the 20th of June 2017.
This event will give you the opportunity to learn, discuss and network with your peers in the MI field. Backdropped in one of Berlin’s most vibrant and artistic locations, break free from traditional conference rooms and share a drink in a typical Berliner Biergarten.
The M.I.E Summit Berlin 2017 will provide you with two in-depth event tracks (keynotes, workshops, and panels) as well as over 25 leading speakers and unparalleled networking opportunities.