Add to Watchlist

Python in the world of retail and mail order


Citation of segment
Embed Code
Purchasing a DVD Cite video

Automated Media Analysis

Recognized Entities
Speech transcript
thank you for joining me on on on the side of the world although we have so my goal is to how realize that the agency used by variance in the way and you for coming in and of itself I will show you we have we have to worry about and all by the using the what the essence of real problems what do you think we have that of the origin of life what do you get there and framework it's so is you and this presentation is also the samples of the reverse so everything going
that OK we have there is a lot of the we want has to make what you are going of so right now the rules of the theory of why that is of the order you get back to you I think we just uh whatever well what was delivered so what do you think OK you can always write among those because it gets assigned value of like and environment like that but and we're talking about about 100 fold into 1 of the items in this this where all the company we thought that he had at the time derivative of good does that really but we only getting an estimated from our customers that about 4 days from order to delivery well there are a lot of silver and hence we got a lot of little that good that quite hard to and will also transforms customer if you order something to keep in mind OK I'm ordering but I'm getting also stuff back with some delay about 1 2 was really depends on on talk 1 I don't want to reach that's why I mentioned supermarket we got some unknown of goods sold every day that they need to build stuff like that and all the the right amount and the united models the situation is out of the things that we're your own challenges and the and have to handle as well as the excess supplies of goods in the evening because know sometimes at Friday we right and then you've got special events like soccer championship take care of everything we talking about something that I think that having occasions so you've got attention you and made all the samples you don't have application only 1 occasion and you've got you should get fast moving goods here because this is the end for and and stuff like that of sold in them all the bond run hundreds star so what what is the essence of the OK you have to do that and you have to know the demand and what was the use of time and in advance and you have to know the frequency of order so but what is the right amount of light and considering the condition of the following conditions you can have to calculate the correct amount I so why would we need in this framework 1 of my talk covers
more or less the prediction predictions simulations replenishment module planted that everything else is necessary but that had not happened over time to we put it into the framework and input output this you should at it yeah you're dealing with companies safe to bring the data to us sometimes it's group sometimes you got problems defining interfaces and also open interfaces are quite difficult sometimes and while I could be done to CC 5 so it's no database and nothing else nothing in so I will talk about for each these but only very rarely note that predicts a slight moving average although we mean and the replenishment I will make some assumptions so the calculations you you can follow while testing I will talk some in the book titles as while making some energy to test for my function values and framework by plotting plotting documentation employment according to monitoring is also very very important to keep in mind if you that we and Singaporean then you've got while minimum set of free for framework OK for prediction well but we need to know tomorrow's demand the message period of time and its attributes we in advance and I the model examples like they think yesterday thinks states some of the sales of same we think well enough that all loading the the moving average you can use gradient was whatever you like from 2nd the articulation of I makes
assumptions OK real itself but got of the stock and content in the evening of 0 so it's growing everything away state because easier to 50 and make a lot of helpful I think that was thrown away each day they don't they don't have to do do with stuff simulations complicated 1 and what about intervals 1 day so everything gets delivered after 1 day 5 years old OK you've got to stop you know the stock and even your you will get to that in order for that so you know OK what was coming in tomorrow and I've got a demand to fulfill those accustomed and while and all that do not tomorrow minus the stop in the evening and the start the ring 0 so the the demand OK for bad reasons I'll explain the data I'm doing a simulation so I predict that the the game of the state and that found all of the cases they get to this demand all I can do it really mean of the last time this and then I although for different strategy for example take the expectation value of the prediction or more complicated take account of the assumed probability so let's say you should can assume the goods are sold by the end of a sentence of using or something that mortgage at the time of course distribution you can calculate this institution for the expectation value you forecast and then take a quantile of a simple reading of OK let's switch over and over and over on simulated data
processing and data generated the repository you can do not going and it produces data are at time series with a gun apostle distribution between 1 and 10 can put in a number of items were allowed to depart the yeah well let's see coming so I'm plotting 1 product so that the whole time series histogram where you see OK we've got some very very low not involved you fail and you've got about 40 % of 1 and so on and while this season for so that's approximated security and means for people and draw from time series of that that is the of simulated much answers
from 2014 FIL and often of results of this in this in the World Series of fluctuating between the 10 more or less and that also the sum of all sets of all I've got a call the and municipal fire so by from presentations state the for each product so my productivity goes from 1 to 100 OK each 1 of them in between 0 and 10 it's going to to home OK well I
would be some the more from simulations but you can't some sort of the strategies like on 1 of them 1 of the expectation value of the quantile make different predictions I can show you what shall
I would like to but some of which are so of the simulation will take about statistics and giving it context you can have a look at this conflict it's a moving average like everything well I on name on you and we run and we can have a look and see
configuration file value of
a system configuration file and giving users program and palette of approximately similar quantiles 10 99 and some steps thanks models and predictions we know 0 OK that's I shall use a corpus of predictions that we assume nothing here you but a start date and end it was the and stimulation and so replenishment would you want to use those of like psychological strategy OK let's see core of the simulation code so those monitoring reading and data frame takes some documents and evaluates the prediction function and if the argument is a prediction function and thus the replenishment and calculates was too much what's missing and
from there so you've got the rate as a result of this information for each product and the each day OK and prediction errors in yet while to right not simple prediction that's just shifted by of sales in my data frame image of the 0 there is and cultural values also 0 putting in and that means OK I haven't the even and I know OK if nothing is done but I know that is the last state for example the fact what is for so my guess from expectation values 5 for tomorrow OK so it's as the run
and because some called you know what 1 like other restaurants because
very yeah and what we call the index and surplus means OK how many items that that sold and what's that must block and it can be because of what it means of also very I've got a a lot of that means OK how many how many days has decided to hold a stock and you know all items so the following content point here getting because I want to get that's close to you ordering and what you've got a lot of that almost so what you can do
it for every of those combination prediction replenishment and you're getting dressed like that those seen or put and well yeah I put on the definition of success rate and almost operate the focus on present and I can think of stock with all items and state and you get a working class depends on the customer wants a customer that's single-case assimilated some expectation values that would be the usual way OK just ordering what had been sold yesterday and getting a value called while I was talking about 50 % then just some excess but we can do better you can take so that really mean for example was window 5 on this is yellowish of hope conceived and you're getting something composed of about 40 % you get about the universe and accessories closest credit would benefit yeah you have the benefit of 6 per cent know what matter was along the can't you have working point for a replenishment of included in in the context of and when you get it OK I did since the summer annotation calls so I didn't write it in very optimized way so is sort of taken from other fields that after what is the use of processing the Library of policies when I got 1st calls on this topic and well with the rest of map function the function the simulation that are an output in the list of values is the values of the allies each stock calculated for 1 quantizer the 1 didn't want and all projects erm of or commonly used in practice for calculating it on multiple holes because you was that same so 650 sold property of the simulated for all 25 quantile and lower part time period of 60 days the tech-support support 1 hour of 90 CPU was come the more complicated it's going to be optimized for speech OK testing but you've
seen what some functions In my
cold in the predicted functions you be functions you of course and dissemination of certain and functions yeah simulation quantile add retrieval tests and this and this just very simple assuming some data frame according to data from the function and the result and using part of
just that and looks forward to his thoughts and you directory what's really of the community and the old policy and I can put in by using also part coverage so I can't say of OK this means so
as to characterize and test for harmony president of problem of the so it's written all needs to have found the remainder the i him to come so I've tested
those directories replenishment information directory and you the coverage of 99 % of the ship and can you can look into it and then you got to and everything's we this function has been tested whether you should you don't have every case but scans
you can test every function here so documentation test by using the fast just for importing nothing in it's similar to test that while I didn't test it is the red was to function to look at the overall because of coverage missing plentiful from OK and I also my conflict while I'm using the images package on private quite nice if you
want to test config file so as
a security of constant conflict of the simulation with diction and i've written metadata runner passes want young constant time testing it the conflict the metadata right here this OK I'm going to end what the bottom part of the object and so on the best documentation is quite as possible to make and you have to find some and this so and the in protecting 1 and then I have lots of text for dates the year came and the simple things those a quantitative in between 1 and 100 the not in the age of 9 9 9 what does this bring to the notice of the system of the data this is a form you know months and in performance the market and yeah
right you I can put for example simulations parents type simulation and nothing happens
all the evidence from exchange of go all German go to the
relations and put in the spring because the window and I I
and run it again and again such I can put it nicely cannot accept OK the expertise involved in data prediction started were instructed to actually before
not it is not the case that was a
very yeah announcement of the
topic because that both of because the testing and thank while replenishment while this thing called the use and the simulation software different you don't stand a lower range of 1 we define the working point from the last cover showed you so yeah I think is important so I have the whole idea artist of radio that is that we must understand 1 of practice that means to us and 66 per cent of those 2 or 3 days the comfortable standard form of empty shell and customer well grocery store or something like that the throwaway about countries we present what is going what you something that's quite high and throwing everything away so that at the worst case scenario you should have gotten in presence doesn't experiments for from all tools we depend on the product so the company has the
same goal and management
and it's just called order of a model of what a prediction model of of finished with famous information there are some were in all of its predictions not that refer them calculating prediction and then of what prediction which is my demand for the next day and I live London on it and the revenge functions and replenishment functions defined benefit would so I can give a into the conflict and say OK all at once all the time and would not good although it's a set of expectation value prediction was more prediction and Europe and then management and during line the process this diffusion put quantum the prognosis for of the this a number there all of that is just that is minus of 0 and the models of calls I'm just going input from all sources of income and some of simplistic custody and all so running for although
there is some sort 15 just want to wonder the put this takes about 5 seconds this so what what we just calculated order and usually you putting that in the database and the new world already putting it on how would find the customer through the cell output files and doesn't automatically those British organization we have to look into
that generator it's called just a simple if you follow there was a they the product of the same that phase must choose is you don't know that which and prediction and all so my prediction sex and I'm wondering about 670 % for long time so that the hierarchical it's about and so we can go some what and after looking into a product of the and models
regions of the everything this is all distributed between 0 and 17 the predictions with no lower that and what's my average all forces that it's 6 the money was only about 6 can as some in 6 of 6 prior to that and my my a friend of a friend the simulated data 1 of the average of 5 so it's not so bad OK the what is missing
I I didn't talk about smoking but it quite important if you're going to be brought to a local stuff going on you want unintelligible that's wrong because the system is usually quite stating that sometimes something happens like customers sense from data of memory which you didn't expect that the of the lands and you cannot so everything is covered by the standard library block and his greatness everything in something like that to make this the lot data and the monitoring problem is very important in because of data see the father nice but I am a facility right so I just want to see the think plot some data and visualized can use different tools like much properly and complements the for evidence from simple questions all used to be when we had some talks around here market documentation is also important I just wrote down in which guaranteed is going to function as a common but we can work with you know who knows you can use things and also if you have got some productive system or testing system you wanted to draw and at the moment you I'm using integral saying this analysis on here and you can just looking at the pictures putting it all together what we have we've got prediction accuracy but management tools put together to the replenishment module and got stopped simulation which is very very simple but it's easy to enhance that and personalization what also would be nice to have this reputation walking monitoring tests and configuration we have already shown you and deployment of reporting reporting the customer wants a the data you want of homing it might items from of much more than that in the words shelves what's the average of the average amount of light and all that stuff like that because this should lead to something so they and put it in the database sometimes it's very hard for our customers to to get to the top so what is the role and and other companies we don't life too much important stuff because it takes time to get data then you should have to look at the data also because as a question with our customers those have that people they accept the rituals 2 of the cell and we say OK we were monitoring reporting now we just made a lot of 8 . 5 % in the sense of what a lot of to process the customer because in the real world and possibilities lies but then double is even better OK but what we have to understand the management framework with the basic
solution for the problem at all for us that in the case every day although every day replenishment and it's quite simple to put this together with the tools of Python standard library and so on and now I give it to you and welcome what's on Friday has code easy and you can use it in everyday life and to to really value for companies because throwing away it's all a shocking delivery times it's very very good for all customers and they're quite thankful for us that you will find that was a good prognosis and forecast thank you and the rest of us all of the this is usual and if you like to ask questions if and the than
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation


Formal Metadata

Title Python in the world of retail and mail order
Title of Series EuroPython 2015
Part Number 128
Number of Parts 173
Author Mack, Philipp
License CC Attribution - NonCommercial - ShareAlike 3.0 Unported:
You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal and non-commercial purpose as long as the work is attributed to the author in the manner specified by the author or licensor and the work or content is shared also in adapted form only under the conditions of this license.
DOI 10.5446/20185
Publisher EuroPython
Release Date 2015
Language English
Production Place Bilbao, Euskadi, Spain

Content Metadata

Subject Area Information technology
Abstract Philipp Mack - Python in the world of retail and mail order At Blue Yonder a lot of different python packages, provided by the community, as well as our own self-written ackages, are used in order to provide flexible solutions to our problems. In this talk I'll present a walkthrough of a generic python application example for demand and purchase order quantity calculations, putting together those packages in an orderly way. The example will feature real world problems derived from hands-on experience with our retail and mail order customers. Additionally the talk addresses the subjects of testing, configuring, parallelising and deploying the code.
Keywords EuroPython Conference
EP 2015
EuroPython 2015

Related Material


AV-Portal 3.5.0 (cb7a58240982536f976b3fae0db2d7d34ae7e46b)


  591 ms - page object