Anomaly Detection : Time Series Talk

[@teqai]

Thank you, I'd be interested in more on anomaly detection.

[@rameshthamizhselvan2458]

Crystal clear and crisp . One of the Excellent Video is what I'm looking for...

[@deter3]

Finished all your time series analysis video in one day and learned a lot !!! Thanks a lot for these videos . You have made complicated things much easier to understand and exploration .

[@hongkyulee9724]

Time series is my head killer. And you are my pain killer. Thank you for the nice video.😀

[@raasheedpakwashi2961]

more anomaly detection methods please!!! brilliant vids btw, been invaluable for me in learning TS Analysis

[@annalelchuk1023]

thank you SO MUCH for your time series videos, both theoretical and code examples. your explanations are brilliant - best tutorial on the topic ever. THANK YOU

[@kanuparthisailikhith]

Extraordinary ❤ … looking forward for more videos on time series irregular

[@overgeared]

nice series. would like to see more on recognition of anomalies and patterns in financial time series such as using HMMs

[@feedbackcontrol]

I just wanted to say thank you. These are excellent videos.

[@ritvikmath]

Glad you like them!

[@zollen123]

Please tell us more about the more robust anomaly detection methods! I need them in my life!!

[@ritvikmath]

great suggestion!

[@sreech9726]

Thank you very much for a wonderful explanation and for sharing the code. One of the best videos I have watched on time series anomaly detection. I would like to learn more about the other robust anomaly detection methods. Could you please share your knowledge on that aswell?

[@shettipellikomal6421]

Yes we need more anomaly detection for time series

[@MrNitroklaus]

This method might work if you know that there is going to be an anomaly. However, using real data you won't know whether or not there is one that (you want to correct for) at all. For this method to be robust, you would have to define rules on when to act at all, e.g. if one of the leave-one-out standard deviations is x% lower than all the others. In practice this is, for example, done by defining a test statistic (in this case the leave-one-out standard deviation or a function of that) and using this statistic to define a Neyman-Pearson type test that decides on whether or not there is an anomaly. And you declare the part of the data that minimizes/maximizes (depending on how the statistic is defined) to be the anomaly.

[@karishmakothari4949]

Really very helpful video. please share more video on anamaly detection

[@anthonyshea6048]

Hey when you calculated the standard deviation which distribution did you assume? I usually assume a poisson distribution for time series data. Is that correct ? Should I use a Gaussian ?

[@saurabhghosh7800]

Very engaging video....

[@Thamizhadi]

Hi, could you please provide resources on how to deal with anomalies that are linked to actual crises rather than data error? The simple method of correcting the anomalies with mean values seem to be appropriate when the anomalies are linked to data error

[@komalkukreja4441]

How to handle multiple anamolies in web traffic forecasting?

[@talktovipin1]

Nicely explained.....please upload more videos....

[@devenderkumar-pr6ig]

can,any please help me for ml project sir,can you halp\\The second topic involves application of time series data analysis - to be specific, detection of anomaly in a set of time series data. The detection may need to apply machine learning based anomaly detection algos. or some simpler algos. (depending on the data). The objective is to detect anomalous driving patterns of a vehicle from its successive GPS positions. In an intelligent transportation system (ITS), a vehicle is supposed to broadcast a periodic message (known as beacon) containing its GPS, current velocity and some other information. The anomalous driving patterns in which we are interested are: drunk driving, aggressive driving etc. for any vehicle, and fraud-route driving for a hired taxi driver (who can take a wrong path intentionally to cheat the passenger). The successive GPS positions of a vehicle, collected from its beacons, is a time-series data, and we aim to detect anomalies in it (using historical data, threshold value etc.). People interested in such types of protocols are: traffic police authorities, / insurance companies (who calculate premiums based on the risk profile of a driver).

[@vignesharavindchandrashekh6179]

Hi.. Thank you for this wonderful video..one question though if we were to detect anomalies for more products then how we should go with anomaly detection rather than doing a plot to find standard deviation by month

[@mohamedelsharawy1138]

Thank you , and i am looking for more on anomaly detection in TS

[@shettipellikomal6421]

What's the difference between change point and an anomaly?

[@bernhardgrusch5035]

First of all thank you for making this video, it's a nice tutorial for getting into this topic! I still have a question about correcting the anomaly. Why did you use the average for predicting the upcomming data? Woundn't it be better to use the median instead, because it's more robust to outliers?

[@umangjain4308]

Thanks for the tutorial.

[@devenderkumar-pr6ig]

can,any please help me for ml project sir,can you halp\\The second topic involves application of time series data analysis - to be specific, detection of anomaly in a set of time series data. The detection may need to apply machine learning based anomaly detection algos. or some simpler algos. (depending on the data). The objective is to detect anomalous driving patterns of a vehicle from its successive GPS positions. In an intelligent transportation system (ITS), a vehicle is supposed to broadcast a periodic message (known as beacon) containing its GPS, current velocity and some other information. The anomalous driving patterns in which we are interested are: drunk driving, aggressive driving etc. for any vehicle, and fraud-route driving for a hired taxi driver (who can take a wrong path intentionally to cheat the passenger). The successive GPS positions of a vehicle, collected from its beacons, is a time-series data, and we aim to detect anomalies in it (using historical data, threshold value etc.). People interested in such types of protocols are: traffic police authorities, / insurance companies (who calculate premiums based on the risk profile of a driver).

[@adityaghosh8601]

Can you please make tutorial playlist for calculas used specifically for machine learning.

[@craigfranze9718]

Could you do a video on a seasonal VARIMA model?

[@mohammedghouse235]

Please make videos on Anomaly detection using KNN.

[@HealthyFoodBae_]

Yes please

[@rtx4070ultrawidegaming]

Robust anomaly detection!!!

[@devenderkumar-pr6ig]

can,any please help me for ml project sir,can you halp\\The second topic involves application of time series data analysis - to be specific, detection of anomaly in a set of time series data. The detection may need to apply machine learning based anomaly detection algos. or some simpler algos. (depending on the data). The objective is to detect anomalous driving patterns of a vehicle from its successive GPS positions. In an intelligent transportation system (ITS), a vehicle is supposed to broadcast a periodic message (known as beacon) containing its GPS, current velocity and some other information. The anomalous driving patterns in which we are interested are: drunk driving, aggressive driving etc. for any vehicle, and fraud-route driving for a hired taxi driver (who can take a wrong path intentionally to cheat the passenger). The successive GPS positions of a vehicle, collected from its beacons, is a time-series data, and we aim to detect anomalies in it (using historical data, threshold value etc.). People interested in such types of protocols are: traffic police authorities, / insurance companies (who calculate premiums based on the risk profile of a driver).

[@alwaaffa]

can you help me with a master’s thesis for my software part (coding) in Python?

[@jackzheninghuang891]

where is the notebook? Thanks

[@rekhasanju938]

Please share Data