Timeseries Modelling using AutoML

[@rameshthamizhselvan2458]

👌Never miss any of your videos.. I was trapped by u r channel there is something unique in it .I was more attracted to the way you put new things that doesn't exist in any of the ML books.

[@AIEngineeringLife]

Thanks Ramesh

[@AkashSingh-ed6vo]

AttributeError: module 'auto_ts' has no attribute 'Auto_Timeseries' Getting the above error!

[@sarathpuppala6148]

Great Video.

[@valerysalov8208]

is automl used in production environment? In your org do you'll have a product or in POC mode?

[@AIEngineeringLife]

We use it to generate baseline models and then once manual features are generated use it to pick hyperparameters and best model again.. It is not used by itself. it is AutoML + Human

[@valerysalov8208]

@@AIEngineeringLife there is feature tool for feature engineering why do manual feature engineering?

[@nehareddysanagala7373]

Sir, I do not see data normalization before feeding it into the network, any particular reason? Thanks for sharing!

[@jialunchen9810]

Hi, thank you very much for your videos. May I ask how do I use your code predict some data without seasonality, so it is not normal datetime.

[@VeereshIttangihal]

Thank you for your effort to share the knowledge this way. Wonderful tutorial. I have got a question to ask, Can I save the best model and use it for predicting the unseen data? I tried to do but I am not able to get new predictions on new unseen data. even though I use unseen data it gives the same output as it gave for the previous test data used while training.. the prediction result is not getting updated based on the newly passed data.. could you please help me to understand the reason or things to do to get it done in the right way?

[@sakilansari9511]

It really helps me.It would be helpful if you provide us code as well which you showed in the tutorial.

[@AIEngineeringLife]

Sakil.. I have uploaded it in my git repo - github.com/srivatsan88/KZbinLI/blob/master/Auto_Timeseries.ipynb

[@abhishekv7171]

How should one approach Multiple Time Series problem, i.e. when you have to forecast 4 weeks out sales for a number of stores? Can such problems be tackled by using LSTM? Or we train using a "for loop" for each store?

[@PallatiCharan]

Easy explanation and great tutorial on automl sir, can you please make video how to apply automl in realtime streaming data(sensors,etc) simulate on static data.

[@AIEngineeringLife]

Sai.. Do you mean deploying AutoML code on streaming pipeline?

[@PallatiCharan]

@@AIEngineeringLife ya yes sir, I have used lstm for realtime streaming but I want to know how an automl can apply in streaming pipline?

[@youcefahmanache3298]

Hi Srivatsan, what is the difference between modelling a time series dataset using Auto ts and AutoML?

[@AIEngineeringLife]

Both are same.. AutoML was created during AutoTS package initial launch and it did not have capability for multivariate series. New video I added on AutoTS contains all latest feature and update to package.. I will re-arrange my playlist sometime over weekend to have only latest version of it

[@youcefahmanache3298]

@@AIEngineeringLife thank you very much for your swift reply, and looking forward to see it. I appreciate and like the work you do. I have a particular interest in timeseries dataset and I find your videos and github repository excellent.

[@AIEngineeringLife]

Thanks Youcef.. Glad if it is helpful 🙏

[@tanb13]

Towards the end of the video, when running the last cell for 'forecast' got a bit confused. I imitated the steps as given in the video still could not run the last cell. Can you please update your notebook on Github with the functionality for 'forecast' and 'fbprophet' added.

[@muhammadzubairbaloch3224]

best lectures I find the link keep continue sir. Thanks

[@manjunath.c2944]

if I run all the steps as specified in the video in Colab for large data set of 2,00,000 rows and 1 column using GPU/TPU its showing "session crashed after using all the Ram "

[@AIEngineeringLife]

Manjunath.. When you are running on large dataset. Change model_type as prophet or ML. Best will run Arima model as well and it will take lot of time to find the right parameter on large dataset

[@manjunath.c2944]

@@AIEngineeringLife thank you very much

[@dipinpaul5894]

Great content. What's your thought/take on AutoML taking up DS jobs in coming years ?

[@AIEngineeringLife]

Dipin.. I have a video on it in case if you have not seen it :) - kzbin.info/www/bejne/iKq3ZWqJoLebo8k

[@dipinpaul5894]

Thanks for the link

[@sanjogh777]

Please don't make such comments and try to scare people. AutoML ain't replacing Data Science. Data Science is more than just using some Data Science algorithms. It's also not just about cleaning data, the hardest part is about understanding the business problems and having the correct domain knowledge to set up a proper framework. You can't automate a domain knowledge just like AI and automation can't replace a mainstream profession like a Doctor or an Electrical Engineer etc.

[@AIEngineeringLife]

@@sanjogh777 Can you be specific where it is said. If you have seen the video I have pasted in above the comment the exact point you have quoted I have highlighted in the video

[@shivanandprasad2611]

please explain softmax activation function in deep learning.

[@deeptigoyal4342]

How can we perform Multivariate time series using Auto ML?

[@AIEngineeringLife]

I do not think currently AutoML support multi variate. I have not tried atleast let me check and update

[@deeptigoyal4342]

@@AIEngineeringLife okay Sure :)

[@vinothkumar-xw1wy]

Great video. Would like to see your time series video with time series components n hw eda need to be handled.. U r rocking👍

[@AIEngineeringLife]

Thank you.. I have planned a series of video on TS. So you should see lot coming on this topic over time

[@aminemairech4100]

Thank you ! link to the notebook + dataset plz ?

[@AIEngineeringLife]

You can find it in my repo below.. datasets must be in dataset folder - github.com/srivatsan88/KZbinLI/

[@불고기-u5o]

Hello. Thank you for the good video. I'm a beginner who just studied machine learning. With this library, can I make multivariate time series predictions?

[@AIEngineeringLife]

Yes you can. This package supports multivariate as well. Latest video on this package is here - kzbin.info/www/bejne/i4O4mol6aJyZpNU You can also check my Time Series course in my channel where I cover multiple ways of doing multivariate time series modeling

[@muhammadzubairbaloch3224]

please sir make more videos on time series analysis using RNN, Arima etc

[@AIEngineeringLife]

Yes Muhammad have it in my plans and u must see it soon

[@khushiyoka_Safarnama]

Hello sir, I followed each of your steps but its showing error like autots has no module named Auto_Timeseries. What can i do?

[@valerysalov8208]

is a data science interview question answering bot a good portfolio project to show for data science interview?

[@AIEngineeringLife]

Yes if it can also parse answers and validate it based on diferent variations of answer

[@valerysalov8208]

@@AIEngineeringLife what do you mean by parse and validate?

[@deeptigoyal4342]

How it works on multivariate time series?

[@navjotsingh6529]

It is a really informative video. I am trying to analyze some sensor data generated at 3 sec i.e. 480 data points in a day. I am not able to set the parameters for Auto_ts to create a time series. I am getting some error while using the parameters in Auto_ts package( seasonal period and time_interval). Any help would be really appreciated.

[@AIEngineeringLife]

what is the error you are seeing ?

[@navjotsingh6529]

@@AIEngineeringLife Thanks, error was fixed. But just need one clarification. Let say our train data index ranging from 1:100 , using AT.Auto_Timeseries(forecast_period=20) will give the forecast value for index - 80:100 and not 100:120 . Is my understanding right ? Also, I am not able to find the way to predict for future time frame using AutoTs

[@Jam05_]

Paaji can you please share how you got overcome with your respective problem?!

[@navjotsingh6529]

@@Jam05_ I was having data at 3 min span , I used frequency = 'H' that solved my problem

[@Abhishekb936]

Hello sir, I am facing issue while installing package. It gives me error 'Failed building wheel for pyflux'.. Could you please suggest here ?

[@AIEngineeringLife]

Abhishek are you building in local or colab?. If local try upgrading pip and see

[@Abhishekb936]

@@AIEngineeringLife thanks sir.. i upgraded my pip in command prompt to 20.0...now it taked forever to run !pip install auto_ts.. I also tried -m pip install auto_ts.. Still it takes forever to run pip command

[@Abhishekb936]

@@AIEngineeringLife yes sir I am.working in local.. Jupyter notebook

[@AIEngineeringLife]

Abhishek you want to better give this a try on colab?. Maybe python version can also be a problem. In colab it works

[@Abhishekb936]

@@AIEngineeringLife Okay sir.. Thanks a lot

[@kruparani8012]

Hi Sir, i tried on local machine in jupyter notebook, but pip install auto_ts /pip3 install auto_ts is not working..Please help

[@AIEngineeringLife]

Can u do a plp upgrade pip and try?. Ideally it must work

[@Jam05_]

Try this sudo pip install [package_name] -U

[@eternity-qy5bv]

I need code Which you used in automl

[@kishorekarunakarapoojary4962]

can you provide this jupyter notebook