Articles, Blog

Time Series Forecasting Theory | AR, MA, ARMA, ARIMA | Data Science

December 16, 2019


In this video I’ll discuss about the theory of basic time series analysis. So we will learn what are the different types of econometrics models In time series forecasting So, we will basically learn what AR, MA, ARMA, ARMA OK? sO,

You Might Also Like

100 Comments

  • Reply Najmuj Sakib December 19, 2017 at 5:56 am

    Helpful. Thanks

  • Reply Sri Gautham December 19, 2017 at 6:06 pm

    At 32:04 you have mentioned that differencing is actually differentiating. Isn't it actually subtracting with the order of the backward shift operator? Say first order differencing is actually yt – yt-1.

  • Reply Deepak Kannan December 27, 2017 at 5:00 pm

    Hi , For Strictly stationary you told mean , variance and covariance are time invariant, that is constant. For weakly stationary also you have mentioned mean, variance should be constant. Can you please help me on this ?

  • Reply Shakeel Ahmed January 2, 2018 at 5:46 pm

    sample plz explaination

  • Reply Daniel Nunes January 7, 2018 at 2:04 am

    I really like your video, it was very enlightening. I'm also looking for some explanation on correlogram and Augmented Dickey- Fuller Test in Eviews analysis, not so much of the programing part itself. I tried to look it up, but I couldn't find any view from you. Is there any?

  • Reply Mehdi Mohamadi January 11, 2018 at 4:10 pm

    it was good. i need some information about time series and sarima models in stata. can you help me?

  • Reply Luis Miguel Delgadillo Soto January 12, 2018 at 1:48 am

    this was the best video on you tube of time series!! I from Perú, thanks!!

  • Reply Abhinandan Nuli January 14, 2018 at 9:41 pm

    man, you could have given the presentation a 100 times better than the current one.. you've made an interesting topic sound like a boring s#%t

  • Reply सिद्धार्थ शर्मा January 17, 2018 at 9:22 pm

    Very Nice..Thankyou

  • Reply Nouer Uz-Zaman January 19, 2018 at 8:00 pm

    for arma model

    ok p denotes to past value..fair enough

    but what does q stand for, so do we just took random alphabet to refer to random error term

  • Reply Priyanka Sreemayi January 24, 2018 at 10:51 am

    Thank you so much Sir 🙂 very well explained

  • Reply Analytics University February 4, 2018 at 9:20 pm

    Credit Risk Analytics Study Pack (PD, LGD, EAD modelling) : http://analyticuniversity.com/credit-risk-analytics-study-pack/

  • Reply amor castromayor February 6, 2018 at 4:52 pm

    Thank You! Big help! 🙂

  • Reply SORAAJ February 10, 2018 at 6:11 pm

    thank you, sir, excellent teaching please share this at [email protected]

  • Reply zenapsgas February 11, 2018 at 6:35 pm

    Where is the death_Age.csv file?

  • Reply Pooja Pushparaj February 16, 2018 at 11:45 am

    Thankyou very much Sir!This has been very useful.

  • Reply abel bruce February 16, 2018 at 6:03 pm

    Good presentation, please kindly share your PPT with me at [email protected]

  • Reply lutfi chandra February 27, 2018 at 4:31 am

    good

  • Reply Tom Jek March 5, 2018 at 12:10 am

    for god's sake, one ad every 5 minutes lesson.

  • Reply Harsha Vardhan March 17, 2018 at 7:41 am

    Regarding white noise series,You have mentioned that mean of the series will be 0 for white noise series and then later told that time series shouldn't be applied and take the average of the series..If mean is 0 how does it matter of calculating average

  • Reply roff poff March 17, 2018 at 5:38 pm

    Top' s my professor's lectures by a mile and a 1/2!!!

  • Reply Smarika K March 20, 2018 at 6:25 am

    At 22:10, for white noise you told that the average is the best forecast. But isn't average = 0, for such white noise time series?

  • Reply Vignesh Arasu April 6, 2018 at 6:00 pm

    Good video. Nice clear explanations. I am taking time series/forecasting class and have tests coming up soon and this was a good summary of concepts. We use SAS in this class, I wish we used R as it is much easier

  • Reply Johana dom April 10, 2018 at 7:26 am

    Thank u¡

  • Reply konduri murali April 10, 2018 at 5:04 pm

    good explanation of time series. But I would like to know which software is used.

  • Reply J Fokianos April 13, 2018 at 8:58 am

    Good tutorial , but is totally ruined by the really hard to understand handwriting.

  • Reply akion xc April 16, 2018 at 3:56 pm

    Good introduction, thanks for sharing !

  • Reply Moin Yaqoob April 21, 2018 at 1:07 pm

    can u please share this ppt at
    [email protected]

  • Reply Samina Tasneem April 26, 2018 at 2:47 pm

    too much ad its boring now

  • Reply SHRIRAM NERKAR May 1, 2018 at 7:05 am

    Lot of distraction by adding zomato ads ….

  • Reply Zhu Yuwei May 2, 2018 at 8:14 pm

    Thank you for your work.
    There are quite some things that can be improved.
    The random drawings on the PPT is a little bit too much and distractive.
    Another place for improvement is the examples given for each slide. There are times the speaker was thinking about examples. It is better to prepare a few examples for slides as a cheat-sheet like note before the speech. Recommend example style from Khan Academic.
    The third place to improve is accent. Though this is a little bit too hard to improve in short time, please at least consider improving it in the future.

    Thank you again.

  • Reply Mamakgooa Mofokeng May 5, 2018 at 10:48 pm

    Very good…

  • Reply Ahmed Iftikhar May 8, 2018 at 12:40 pm

    Great !

  • Reply Akin O. May 8, 2018 at 9:17 pm

    Holy cow! Ads every 3 mins! Jeez!

  • Reply Sowmya Lakshmi May 13, 2018 at 6:07 am

    One small question. For white noise if the mean should be zero. How come the best forecast is Average? Couldn't understand the usage of mean and average here?

  • Reply Sarah Kashif May 13, 2018 at 11:20 am

    Can I get these slides?I am studying financial econometrics and this video is very useful.

  • Reply Smitha Us June 12, 2018 at 6:17 pm

    Could please do a video on NSE stock exchange data usage moving average

  • Reply Niharika Sai Krothapalli June 13, 2018 at 3:01 pm

    watch the video at a speed of 1.25

  • Reply Edwin Varghese June 14, 2018 at 10:43 am

    The best explanation of time series. It's totally worth the 53 mins. However I have a doubt, when he said about the error in the MA method. Equation what you meant was: xty = slope*xt + intercept. and the error = xty – xt. Kindly correct me if I am wrong.

  • Reply Ali Farahani June 28, 2018 at 1:38 am

    That was amazing and useful. Thank you. I wish I knew what resources are used in this video.

  • Reply Deepti Saxena June 28, 2018 at 10:32 am

    Time series analysis is one of most complicated topics for me and your video has simplified it so much for me. Your handwritten notes also helped a lot to understand the concept. Thank a lot for explaining in such a brilliant way!!!

  • Reply Susan St. James July 12, 2018 at 7:09 am

    Spikes decay toward zero; spikes cut off to zero; I wish that the meaning of this two expressions was made explicit.

  • Reply Abraham Rodríguez August 3, 2018 at 10:49 am

    BEST VIDEO EVER THANK YOU

  • Reply Santosh Venkataraman August 6, 2018 at 9:08 am

    Very easy to grasp and very well explained. Thanks a lot for the video!!

  • Reply aboctok August 16, 2018 at 7:35 am

    Good grief, as Charlie Brown would say. Wayyyy too much "ahh, ah,, ahhh" for my pitifully low tolerance. In other news, I see that his gf or sister has left a superlative comment below. The best on YouTube. Wow.

  • Reply Somnath Mahato August 23, 2018 at 9:04 pm

    Are you an odia guy?

  • Reply krishnachouhan September 13, 2018 at 7:39 pm

    For a particular dataset, i found (p=0, i=1, q=2) and (p=0, i=1, q=3). In both cases i shifted the predicted values by -1 timeperiod.(what i predicted for feb, i conisdered it for jan). Doing so i got testing accuracy of 99% (MAPE-mean abs percetage err).
    What went wrong with this scenario because 99% seems way to good to be correct for any scenario.

  • Reply ppgerberm September 14, 2018 at 5:20 pm

    There are plenty of videos on youtube. If you don't like how he teaches, go find another one. That is the beauty of internet. For me, it served the purpose and now I understand time series. Thank you, very much.

  • Reply ANAND RANJAN September 18, 2018 at 7:22 am

    awesome explanation…!!

  • Reply Narendra Rana October 8, 2018 at 11:19 pm

    I am new to time series,  your video was helpful, it had all the concepts in one place and it was an hour well spent. However  I don't think differencing and differentiating are the same thing. Differencing is the process of taking a difference of series from itself as various lags. I am not sure if its about taking a derivative (like you hinted) …

  • Reply Ritwick Dutta October 11, 2018 at 9:34 pm

    I have shared the video with my class mates. Thanks for putting this up. Where can I get to buy the package? Do you for a student price?

  • Reply WASSCE Tutorials October 16, 2018 at 2:21 pm

    This truly is the best video on the subject. I have regression and time series exam this Saturday! This was a huge help!

  • Reply Hari Basnet November 1, 2018 at 10:15 am

    Thank you so much sir. Your explanation has taught me enough about the time series model. Thank so much from Bhutan.

  • Reply maha ahmed November 2, 2018 at 10:32 am

    does anyone know anything about varima model?

  • Reply gourav Kumar November 3, 2018 at 8:49 pm

    this is really a good explanation of time series analysis for beginners .i learn a lot from this video..thank u very much sir

  • Reply Maker Shaker Waker November 5, 2018 at 12:52 am

    21:15, why would variance be constant?

  • Reply mehraj u din bathangi November 13, 2018 at 7:52 pm

    Great video on timeseries modelling.

  • Reply TheIsrraaa November 15, 2018 at 8:53 pm

    Lol WTF with the adds, here is your dislike!!

  • Reply Farah Ysebaert November 24, 2018 at 9:02 am

    It took me too many minutes (over an hour) to watch this, just because of all the advertisements. Statistics is already not so fun for me, but you're making it worse in this way! This video is not workeable at all anymore. It's a shame, because you explain well in between the advertisements.

  • Reply Rafiullah khan November 26, 2018 at 8:12 am

    good presentation thank u

  • Reply Vishwaas Hegde November 27, 2018 at 5:37 am

    Differentiated once is said to be integrated by order 1? That is really confusing

  • Reply Pavan Kumar November 30, 2018 at 10:52 pm

    By definition I time series has regular time intervals but how do I work on time series with irregular time intervals like sensor data coming at irregular time? Can I still use ARIMA, ARMA, MA models ?

  • Reply Dr Manoj Kumar Mishra December 7, 2018 at 9:08 am

    your lecture is good and clear

  • Reply tisa maria antony December 12, 2018 at 3:54 pm

    One of the best video for learning econometrics basics… Thankyou soomuch.

  • Reply Aditya Rjl December 15, 2018 at 6:39 am

    best explanation for time series .. Thank you very much sir

  • Reply Chen Bai December 30, 2018 at 7:03 am

    what is difference between ARMA and ARIMA?

  • Reply Alexander Barnard December 31, 2018 at 1:29 pm

    Please please please declare your values. Is beta a constant? integer? real? is Phi a constant? It would be helpful to use proper mathematical notation to make it clear!

  • Reply Kavitha Murthy January 25, 2019 at 6:13 am

    A very easy and simple way of time series concepts explanation.. Great job

  • Reply Shamsher Alam February 17, 2019 at 7:27 pm

    Sir, Your way of teaching awesome

  • Reply Baktash Jami February 17, 2019 at 8:35 pm

    Well-done! Very professional, informative, with a lot of contents, and answers almost every possible questions. In the future videos, please also explain specific letters. For example, if you say that "p". or t, or u, or whatever, please explain that those specific letters refer to. Thanks a lot. I took a lot of useful notes.

  • Reply Dipali Bhatt February 26, 2019 at 4:01 am

    Thank you sir…!!

  • Reply Balaji Doraibabu March 10, 2019 at 3:09 pm

    Nicely done! Throughout this session, you have articulated clearly why we need to perform certain steps. That was quite helpful for me.

  • Reply MARWEL WRITING March 18, 2019 at 6:43 pm

    https://youtu.be/Vih2Bw-aEbk
    Time series analysis

  • Reply Анонимуси March 24, 2019 at 7:15 pm

    If anyone is interested, I wrote some Javascript that does this: https://github.com/8483/time-series

  • Reply Mahfoud DELAL March 30, 2019 at 1:38 am

    the presentation is not sexy but the explanation are very good. Thanks

  • Reply michaeltruthspeaker April 24, 2019 at 7:34 am

    Amazing ! Thanks !

  • Reply Murali Krishna Hari May 6, 2019 at 1:20 pm

    Good attempt

  • Reply milimo mashini May 13, 2019 at 3:08 pm

    you save my time …thanks alot

  • Reply Luisa Krawczyk May 18, 2019 at 4:06 pm

    Great video, very concise!

  • Reply Valerie Chen May 21, 2019 at 7:12 pm

    thank you for making this video.. i hope i will ace my exam tomorrow

  • Reply Fahraynk May 26, 2019 at 1:23 am

    You are my hero.

  • Reply Ömer Giray Özay June 3, 2019 at 8:15 pm

    You save my ass for my Bachelor Thesis. Thank you very much !

  • Reply Kishlay Shukla June 5, 2019 at 6:14 pm

    This is the best video lecture for time series analysis on youtubr

  • Reply Amar Jadhav June 17, 2019 at 6:56 am

    Ek number. Best & crystal clear explanations. Keep up the good work.

  • Reply Saurabh Choudhary June 17, 2019 at 11:38 am

    Super theoretical not very intuitive

  • Reply Karishma Saikia June 21, 2019 at 4:49 am

    Thank you so much sir!! So far the best video on time series

  • Reply 125zakir June 30, 2019 at 2:48 pm

    Very well explained, must say.

  • Reply Munish Kohli July 18, 2019 at 11:27 am

    Hello,

    I am new to Forecasting specially ARIMA/ETS techniques. I have watched most of your videos on ARIMA/Time Series and like the way you explain in a simple and slow manner. I am still not able to fully grasp concepts and have lots of lots questions around ARIMA. I started doing some simple programming in R for ARIMA models. Attach small R script predicts US GDP for 2017 and 2018. I did try through both AUTO ARIMA and non Auto. Through non Auto ARIMA my error difference between Actual and forecasted are lower than AutoARIMA. Does this mean Auto Arima should not be trusted. Below is what I explain in R script. I will really appreciate your response. Thanks

    #Actual 2017 and 2018 GDP Numbers

    Year Actual

    2017 59,928

    2018 62,641

    #Forecast with AutoArima

    Point Forecast Lo 80 Hi 80 Lo 95 Hi 95

    2017 58824.37 58170.17 59478.56 57823.86 59824.87

    2018 59790.69 58559.33 61022.05 57907.49 61673.89

    #Forecast with Arima order 1,1,1

    Point Forecast Lo 80 Hi 80 Lo 95 Hi 95

    2017 58987.01 58273.22 59700.79 57895.37 60078.64

    2018 59972.21 58536.88 61407.54 57777.07 62167.36

    Below is my R script
    installed.packages("WDI")

    library("WDI")

    library(forecast)

    gdp<-WDI(country = c("US"), indicator = c("NY.GDP.PCAP.CD"),start=1960,end=2016)

    names(gdp)<-c("iso2c","country","GDPPerCap","year")

    head(gdp)

    #Order gdp in ascending order

    gdp<-gdp[order(gdp$year),]

    head(gdp)

    plot(GDPPerCap~year,data=gdp)

    #This ts function will transform data.frame into sequence of data record

    us<-ts(gdp$GDPPerCap,start = min(gdp$year),end=max(gdp$year))

    us

    #Run Dickey fueller test

    adf.test(us,alternative = "stationary")

    #Run Auto correlation and Partial correlation function

    acf(us)

    pacf(us)

    #Run Auto Arima

    usOPT<-auto.arima(us)

    usOPT

    #Run Manual Arima

    usOPT <-arima(us,order = c(1,1,1))

    usOPT

    #Question: Why manual setting order of 1,1,1 predict forecast for 2017&2018 with less error difference than auto arima which producess 2,2,2?

    #See below results

    #Coefficent

    coef(usOPT)

    predict(usOPT,n.ahead = 2,se.fit = T)

    #Forecast

    GDPForecast<-forecast(object = usOPT,h=2)

    GDPForecast

    #Actual 2017 and 2018 GDP Numbers

    Year Actual

    2017 59,928

    2018 62,641

    #Forecast with AutoArima

    Point Forecast Lo 80 Hi 80 Lo 95 Hi 95

    2017 58824.37 58170.17 59478.56 57823.86 59824.87

    2018 59790.69 58559.33 61022.05 57907.49 61673.89

    #Forecast with Arima order 1,1,1

    Point Forecast Lo 80 Hi 80 Lo 95 Hi 95

    2017 58987.01 58273.22 59700.79 57895.37 60078.64

    2018 59972.21 58536.88 61407.54 57777.07 62167.36

  • Reply Mangireesh Potnis August 24, 2019 at 4:18 pm

    fuck udemy

  • Reply kik kaka September 3, 2019 at 7:02 pm

    92% discount #coupon #udemy #course for

    Complete #Time Series Data Analysis #Bootcamp In R

    #deal Ecoupon

    https://www.udemy.com/course/complete-time-series-data-analysis-bootcamp-in-r/

  • Reply நேரம் தொலைக்காட்சி channel September 11, 2019 at 9:05 am

    Lot of story

  • Reply vikrant bhattacharjee September 15, 2019 at 2:45 pm

    22:00 mins – If white noise is purely random, then how come it has a 0 mean and a constant variance? A specific mean and a constant variance makes a term determinable or predictable, right? Kindly help me on this. Thank you 🙂

  • Reply Shubham Kumar October 13, 2019 at 8:56 am

    Thanks for making this video tutorial. Many of my doubts are now clear.

  • Reply manoj yadav October 16, 2019 at 6:45 am

    good explanation

  • Reply Kabineh Kpukumu October 29, 2019 at 7:24 am

    what an incredible tutorial! Please keep posting many more. Thanks and God bless you.

  • Reply archana reddy November 7, 2019 at 9:19 pm

    What if both ACF and PACF have spikes cut off to zero? Is that possible? Because I am running time series on R and my graphs are showing cut off to zero for both ACF & PACF.

  • Reply farzi janta November 9, 2019 at 7:06 am

    How can I download the dataset used in case study i.e. death_Age.csv

  • Reply Mulunesh Lamore November 30, 2019 at 12:48 pm

    It is very nice presentation style. It help students to understand what main points should be understood.

  • Reply yuthpati rathi November 30, 2019 at 5:17 pm

    Amazingly explained sir

  • Reply Analytics University December 15, 2019 at 5:29 pm

    Data Science Training : http://bit.ly/2PMVRPV

  • Leave a Reply