Building a basic Model for Churn Prediction with KNIME

[@djibb.7876]

Nothing better than that to start with Knime. Excellent

[@balkrishna83]

Thank you for showing deployment process too. 🙏🏻

[@bintorys]

Pre-requisites before the start of the video can be added to help viewer know their level of understanding. Well explained Thank you.

[@kavinyudhitia]

Thanks! This is a good training for a beginner like me! I even learned other thing beside the main topic of using decision tree such as coloring the table and exporting result to reporting environment

[@invdata8326]

Спасибо, очень полезное видео. Только благодаря вам что-то начинает доходить. Thank you

[@hizokadarkwolf]

this was a great introduction, thank you very much. Now that I understand the flow, I'm going to explore other options available

[@sandeepmandrawadkar9133]

Precice and simple explanation 👌 Thanks for your efforts in envisioning 👍🙏

[@TheSpinninHead]

Very well taught. Thanks for this video. Perhaps the first Knime video I understood properly. Could u plz make one for either neural network or pattern recognition?

[@KNIMETV]

Sure, we are working on it.

[@britboss7290]

@@KNIMETV what is the extension where you can get the reporting environment ?

[@diptinemade3279]

Very well articulated and concise trainig. Thank You

[@imtiaziqbal3041]

Great content. Thank you for your effort

[@Drkalaamarab]

Excellent demonstration madam. Thank you from india 👍. More examples in knime please

[@ratansunderwatergopro8203]

excellent video... clearly articulated. Thanks so much

[@mshparber]

Excellent! Thank you for this explanation!

[@Evnovacion]

Great tutorial! Very clear, many thanks!

[@TS-ml4dp]

This is fantastic!!!

[@pranatim]

Very much informative and thank you for sharing.

[@phichayaphakphiphitphatpha3695]

Thank you, it's very helpful for me.

[@MartinaHertl]

Amazing video! Thank you a lot!

[@OurStoryz123]

Thank you very much for this excellent tutorial

[@tinchomrv]

Very useful. I need to apply the same, BUT with time series. I want to predict if a client is going to churn next month, based on its previous monthly historic data. I think the solutions should be similar but involving time series. Not sure exactly on how to combine both. Could you provide some guidance? Thanks.

[@vlera4198]

best explanation what i ever seen

[@JamieJameJame]

I haven't started using KNIME yet, but have watched a few tutorials. I assume the output is a reporting dashboard? It would be nice to see what the output looks like.

[@SB-vj5sn]

Excellent - and so well articulated, thank you!

[@fsalam]

Wonderful presentation! I have a question. All examples that I have seen uses variables in columns. What would be the right approach in a retail scenario where there are more than 10K products(SKU)? You cannot have 10K columns. Ideally, these product purchase transactions should be 2 or 3 variables ( name , quantity, unit_price). Lets assume in this scenario that the product plays a role in the churn behavior. Would a 3 column(variable) approach be suitable in this scenario?

[@rosariasilipo_knime]

Sure. You can use any variable type as input.

[@britboss7290]

what is the extension where you can get the reporting environment ?

[@MK-pf9hs]

Very useful! Is there a possibility to identify the factors that influenced an individuals predicted churn-probability?

[@janOverwatchGM]

YES, OPEN VIEW DECISION TREE LERNER U CAN SEE FOR EXPAMPLE CUSTOMER WITH HIGH DAY CHARGE [for example 44,805] WILL CHURN with probability of 60,5 AND CUSTOMER WITH DAY CHARGE LESS THAN 44,805 will only CHURN WITH AN PROBABILITY OF 11% HERE U GO LITTLE LATE

[@Ps3thi]

JPMML continuously showed errors in my workflow, idk why it wouldn’t work on my dataset :(

[@jlgabrielv]

Great video, very helpful and well explained, thank you!

[@MAbdullah47]

where we can find the resources of this Videio?

[@erlcugnaga1065]

Where is the speaker from ?

[@AdiBandaru]

Thank you for very good demo.

[@ravivissa]

Thanks for sharing the video. It very well articulates about how to build a simple model and I think this model will be used in general by any organization. One question - I am trying to re-create this model for another data set and am unable to see options as per attribute variable. Let me elaborate, for attribute variable, typically the options shown are 1(and we choose a color) and 0(and we choose another color). I am unable to see any option for me to select. Can you let me know where I am going wrong. Well, the coloring apart, I was able to build the model, which is good part, but as you said in the video....coloring, why not? :-)

[@KNIMETV]

Are you using the Color Manager node?

[@ravivissa]

KNIMETV yes

[@KNIMETV]

If you open the configuration window of the Color Manager node you can select the column to use for coloring. If it is a binary column then you get 0 and 1. But if it is another column type, you will get a list of possible colors (if a nominal column) or a heatmap (if a numerical column). If this explanation does not help, maybe you should ask this same question on the KNIME Forum www.knime.org/forum.

[@CEOPrestus]

I'm very excited with KNIME. In my case, the churn can occur monthly, is is possivel to have KNIME looking in the TIME-evolution of the last 6 months? Perhaps via Recurrent Neural Networks (RNN) / LSTM? Thanks a lot, KNIME Education!

[@rosariasilipo_knime]

Yes, it is possible. We are preparing a blog post on using LSTM on text. I think it could be easily extended to churn data as long as time series are available.

[@ithinkib248]

is each column in the combined dataset (produced in node 3 - joiner) a variable that this model is using? how you do specify which ones are dependent and independent?

[@rosariasilipo_knime]

I use all features as input here and select Churn as Target in the configuration window of the decision tree learner node. Some Learner nodes have an Include/Exclude framework to include/exclude columns from the input variable set. If the Learner node you are using does not have that, you can always use a Column Filter node before the Learner node. The configuration window of the Learner node always allows to select the Target variable.

[@SB-vj5sn]

Wonderful presentation . I would like to improve performance by testing correlations, functionally transforming some of the predictors (attributes), and eliminating a couple based on their importance. I am new to KNIME. Can someone point me to a video or paper on how to perform these steps pre- or post the Joiner Node step? These techniques would of course carry over to other models.Thanks.

[@rosariasilipo_knime]

We do not have new material to detect the importance of input attributes. You can check this whitepaper at page 23 files.knime.com/sites/default/files/inline-images/internet_of_things_with_knime_final1.pdf The workflow can be downloaded from files.knime.com/sites/default/files/inline-images/IoT_group_workflows.zip and data from www-cdn.knime.com/sites/default/files/Raw%20Data.zip

[@nibinjoseph3725]

Where is the dataset?

[@sediqkhan8353]

I could not find JPMML predictor in the current version of KNIME, so what is the alternative to it. Because now, the probability of this instance becoming a churn in my workflow is zero "0".

[@rosariasilipo_knime]

The JPMML classifier is part of the PMML extension. You can also use the predictor node fitting your machine learning model.

[@britboss7290]

@@rosariasilipo_knime Hi, what is the extension where you can get the reporting environment ?

[@rosariasilipo_knime]

@@britboss7290 in the Report Designer extension.

[@britboss7290]

@@rosariasilipo_knime I have it but still can't find the reporting environment

[@rosariasilipo_knime]

@@britboss7290 Check this lesson from the e-learning course www.knime.com/knime-introductory-course/chapter4/section3/export-data-into-birt-report

[@0568raju]

Hi it's so helpful and the knime is awesome and very easy to use A question: you have used all the other independent variables to predict Churn but what if there is correlation exists among independent variables or between dependent and independent variables? can you please share the same dataset used inthe video to practise?

[@KNIMETV]

+raju goud M The dataset is available in the EXAMPLES server accessible through the KNIME workbench. It is in 050_Applications/050018_ChurnPrediction. About variable correlation, this is the topic of another video probably! For now you can read about removing correlated attributes in this whitepaper "7 techniques for dimensionality reduction" www.knime.org/files/knime_seventechniquesdatadimreduction.pdf

[@0568raju]

+KNIMETV Thanks and any video on dimensionality reduction techniques using knime available in youtube?

[@KNIMETV]

No video on dimensionality reduction on KZbin, but a whitepaper on the KNIME web site www.knime.org/files/knime_seventechniquesdatadimreduction.pdf

[@KNIMETV]

No video on dimensionality reduction. Just a whitepaper www.knime.org/files/knime_seventechniquesdatadimreduction.pdf

[@sergeykurk]

what's the use of data scientists if we have knime?

[@jeanpaultrinidad1664]

Excellent tutorial.

[@marekbodzianowski3944]

Great Video ) thanks for that , but i have one important question . How can we use the model ( build diagram ) to get answer about churn to each number from the input file. I would like to get answer like this : number | Chur ( prediction) | churn probality 123454 1 90% 234556 1 0% 434444 1 5% 232323 1 95% and so on ....

[@rosariasilipo_knime]

Any predictor node in KNIME has an option in the configuration window to output probabilities. This is the option you should use.

[@mariaceciliagarcia8713]

Excellent tutorial! I would like to know how to add nodes because I haven't JPMML in my version of KNIME . Thank you. (Sorry for my English!)

[@rosariasilipo_knime]

You can use a Decision Tree Predictor node or a PMML Predictor node. For a tutorial on KNIME you can check the Learning Hub www.knime.org/learning-hub

[@shaunaka6557]

Nice demo!

[@marcc2689]

Great video thanks!!!

[@UmdaIT]

Great tutorial!!

[@ianx0114]

Is this dataset available for download? Thanks.

[@ianx0114]

Actually, just found it in case anyone is also wondering - www.knime.org/knime-applications/churn-prediction

[@vlera4198]

catch www.knime.com/nodeguide/applications/churn-prediction/training-a-churn-predictor

[@laraibashqeen7932]

not able to find .csv file bro. help me out here. the second link which have given above doesn't giving me an option to download this CSV file.

[@nongshim881]

Amazing

[@orlandomarcelovazquezlopez1778]

👏🏼

[@HeliosAI56]

sembra ala voce di un italiana, dalla cadenza, dico bene ?grazie cmq del prezioso video, buon lavoro

[@KYC_life]

Why learn Python, R, ... if KNIME does this?

[@AD-dz4un]

learn Statistics, not tool. Knime, R, Python is just for machine learning, if you can not understand the very basics of statistics, then nothing will help you

[@KYC_life]

Aija Daina Statistics is prerequisite to understanding all the concepts mentioned in the video. My question wasn't related to statistics but the fact that considering writing codes in R or Python to perform churn analysis while proper understanding of the concepts can be done faster on KNIME. Thanks for the response.

[@AD-dz4un]

Custom code gives you more flexibility compared to Knime, SPSS Modeller etc. All graphical interfaces have a lot of limitations code wise, so we can use standard methods in our analysis, if more complex equations, models ar eneeded, then all these graphical tools are quite useless. So far , Knime looks very good, first I saw it in 2.1 i've done lot of things with R, so Knime allows me to optimize some part, say 60%, not more.

[@TheSimoyw]

great

[@sandeepm625]

nice

[@HarshSingh-qh4qh]

I love u and love your accent more. Produce video on clustering. Espanol have Glasshour figure.

[@aegystierone8505]

she has Melania Trump's accent

[@rosariasilipo_knime]

it is Italian.

[@fakhrijunaid]

What is the extension where you can get the reporting environment?