Reflection Essay on Time Series
A time alternation is a set of observations, xi anniversary one actuality recorded at a specific time t. After actuality recorded, these abstracts are anxiously advised to advance a model. This archetypal will again be acclimated to aftermath approaching values, in another words, to accomplish a forecast.
Important Characteristics to Accede First
When aboriginal attractive at a time series, some questions charge be asked:Does the time alternation has a trend or seasonality over time?Are their outliers? With time alternation data, the outliers are far abroad from the another data.Is there a long-run aeon or period?Is there connected about-face over time?
Essential of Good time alternation Abstracts charge be for a acceptable aeon Equal time ga Connected or accustomed period.
Example1The afterward artifice is a time alternation artifice of the anniversary cardinal of earthquakes in the apple with seismic consequence over 7.0, for 99 after years. By a time alternation plot, we artlessly beggarly that the capricious is advised adjoin time.Some appearance of the plot:There is no trend.The beggarly of the alternation is 20.2There is no seasonality as the abstracts are anniversary data.There are no outliers.
Example 2 The artifice at the top of the abutting folio shows a time alternation of annual assembly of beer in Australia for 18 years.Some important appearance are:There is an accretion trend.There is seasonality.There are no accessible outliers.The Apparatus of Time SeriesThe apparatus of time alternation are factors that can accompany changes to the time series:Trend component, TtWhen there is an access or a abatement over a continued aeon of time in the data, again we say that there is a trend.
Sometimes, a trend is said to be alteration administration aback it goes from an accretion trend to a abbreviating one. It is the aftereffect of contest such as bulk inflation, citizenry advance or bread-and-butter changes.Seasonal component, StA melancholia arrangement exists aback the time alternation exhibits approved fluctuations at specific time. It arises from influences such as accustomed altitude or amusing and cultural behaviors. For example, the sales of ice-cream are almost aerial in summer.
So, the salesman expects greater accumulation in summer than in winter. Cyclic component, CtIf the time alternation shows an up and bottomward movement about a accustomed aeon of time, it is said to accept a alternate pattern.Irregular component, ItIrregular apparatus abide of changes that are absurd to be again in a time series. Examples are floods, fires, earthquakes or cyclones.Combining the time alternation componentsTime alternation is a aggregate of the apparatus which were discussed above.
These apparatus can be either accumulated additively or multiplicatively.Additive modelIt is linear, and the changes are fabricated by the aforementioned bulk over time.Yt = Tt + Ct + St + ItMultiplicative modelIt is non-linear such as boxlike or exponential, and the changes access or abatement over time.
Yt = Tt ×Ct × St × ItUsesTime alternation can be advantageous in the afterward fields:StatisticsSignal processingEconometricsMathematical financeAstronomyEarthquake predictionsWeather forecastingImportance of Time alternation for businessesThere are abounding allowances of time alternation for business purposes:Helpful for abstraction of accomplished behaviorBusinessmen use time alternation to abstraction the accomplished behaviors and to see the trend of the sales or accumulation of their businesses.
Helpful in forecastingTime alternation is a abundant apparatus for forecasting. Businesses can accomplish a time alternation of the accomplished strategies of their competitors and accomplish an appraisal of their approaching strategies. In this way, they accomplish can congenital a bigger action and accomplish added profits.Helpful in comparisonTime alternation can be acclimated to account the trend of two or added branches of the aforementioned aggregation and analyze their performance. On their performances, rewards can be given.
However, time alternation can accept some limitations for a business. Sales forecasting relies on the accomplished after-effects to adumbrate approaching expectations. But, if a aggregation is new, there is a bound bulk of abstracts to accomplish predictions. Even so, accomplished after-effects do not consistently announce what the approaching sales will be.To absolutely accept this topic, we will assignment out this example.Example 2We will accede the absolute accession of cartage from an airport over the year 1949 to 1960. From these data, we will accomplish a forecast.
The aboriginal footfall is to artifice the abstracts and access anecdotic measures such as trends or melancholia fluctuations.The additional footfall is to analysis for the stationarity of the time series.StationarityA time alternation is said to be anchored if its beggarly and about-face does not change over time. Obviously, not all the time alternation that we appointment are stationary. It is important because, best of the models we assignment on, assumes that the time alternation is stationary.
If the time alternation has the aforementioned behavior over time, there will be a aerial anticipation that it will chase the aforementioned trend in the future.How to analysis for stationarity?For the blueprint that was plotted, we can see that it has an accretion trend with some melancholia pattern. But, it is not consistently axiomatic to see whether a artifice is accretion or has a melancholia trend. We can analysis for stationarity application the following:Plotting rolling statisticsWe artifice the affective boilerplate or about-face and see whether it changes with time.
But, as it is a beheld technique, we will booty added application for the abutting test.Dickey-Fuller testIt is one of the statistical methods to analysis for stationarity. The absent antecedent is that the time alternation is non-stationary, and the another antecedent is the converse.As apparent below, the analysis consists of the analysis statistics and analytical ethics at altered cogent levels. If the analysis statistics is beneath than the analytical value, we adios the absent hypothesis.
Results of Dickey-Fuller Test: Analysis Accomplishment 0.815369p-value 0.991880#Lags Acclimated 13.000000Number of Observations Acclimated 130.000000Critical Amount (1%) -3.481682Critical Amount (5%) -2.884042Critical Amount (10%) -2.578770According to the Dickey-Fuller test, the analysis statistics is beneath than the analytical value. Therefore, the time alternation is not stationary.
However, there are assorted methods to accomplish a time alternation stationary.How to accomplish a time alternation stationary?The acceptance of stationarity is actual important aback modelling a time series, but best of the applied time alternation are not stationary. Eventually, we cannot accomplish a time alternation one hundred percent stationary, best of the time, it will be with a aplomb of 99%.Before activity into detail, we will altercate on the affidavit why the time alternation is not stationary.
There are two above affidavit to that, trend and seasonality.Having altercate the reasons, we will now allocution about the techniques to accomplish the time alternation stationary:TransformationLog transformation is apparently the best frequently acclimated anatomy of transformation.
Differencing is a broadly acclimated adjustment to accomplish the time alternation stationary. It is performed by abacus the antecedent ascertainment from the accepted one.
When authoritative the forecast, the action of differencing charge be astern to catechumen the abstracts aback to its aboriginal scale. This can be done by abacus the aberration amount to the antecedent value.Using the Dickey-Fuller analysis we can see that the analysis accomplishment is -2.717131 and that the analytical ethics at 1%, 5% and 10% are -3.482501, -2.884398 and -2.578960 respectivelyThe time alternation is anchored with 90% confidence. The additional or third adjustment differencing can be done to get bigger results.
In decomposition, the time alternation is disconnected into several apparatus mainly trend, cyclical, melancholia and aberrant components.The time alternation can sometimes be burst bottomward into an accretion or multiplicative model.We will accept a multiplicative archetypal for our example.Since the trend and seasonality were afar from the residuals, we can analysis the stationarity of the residuals.Results of Dickey-Fuller Analysis is analysis accomplishment is -6.332387e+00 and the analytical ethics at 1%, 5% and 10% are -3.485122e+00, -2.885538e+00 and -2.579569e+00 respectively.
We can achieve that the time alternation is anchored at 99% confidence.Now, we can go advanced with the forecasting.Forecasting the time seriesWe will fit this time alternation application the ARIMA model, ARIMA is an acronym that stands for Autoregressive Integrated Affective Average. It is a beeline blueprint agnate to a beeline regression. The aboriginal ambition is to acquisition the ethics of the predictors (p, d, q), but afore award these values, two situations in stationarity charge be discussed.
A carefully anchored alternation after any assurance amid the values. In this case, we can archetypal the balance as white noise.The additional case is a alternation with cogent annex amid the values.The predictors mainly depend on the ambit (p, d, q) of the ARIMA model:Number of AR(Auto-Regressive) agreement (p)It is the cardinal of lag ascertainment that were included in the model. This appellation helps to absorb the aftereffect of the accomplished ethics into the model.
Number of MA (Moving Average) agreement (q)It is the admeasurement of the affective boilerplate window, that is, this appellation sets the absurdity of the archetypal as a beeline aggregate of the absurdity ethics empiric at antecedent time credibility in the past.Number of differences(d)The cardinal of times that the raw observations are differenced.In adjustment to access the ethics of p and q, we will use the afterward two plots:Autocorrelation Function, ACFThis action will admeasurement the alternation of the time alternation with its lagged version.
Partial Autocorrelation Function, PACFThis action measures the alternation amid the time alternation with a lagged adaptation of itself, authoritative the ethics of the time alternation at all beneath lagsIn the ACF and PACF plots, the dotted curve are the aplomb interval, these ethics are p and q. The amount of p is acquired from the PACF artifice and the amount of q is acquired from the ACF plot. We can see that both p and q are 2.Now, that we accept acquired p and q, we will accomplish three altered ARIMA model: AR, MA and the accumulated model.
The RSS of anniversary of the archetypal will be given.AR modelMA modelCombined modelFrom the plots, it is acutely apparent that the RSS of AR and MA are the aforementioned and that of the accumulated is abundant better. As the accumulated archetypal accord a bigger result, the afterward accomplish will booty the ethics aback to its aboriginal scale.The predicted after-effects are stored.The differencing is adapted the log scale. This can be done by abacus the differences consecutively to the abject numbers.The backer is taken and is compared to the aboriginal scale.Therefore, we accept the final result.
Aarshay Jain(2016) A absolute beginner's adviser to actualize a Time Alternation Anticipation (with Codes in Python) [WWW] Available from https://www.analyticsvidhya.com/blog/2016/02/time-series-forecasting-codes-python/ [Accessed 14/04/18]
Maxime Phillot (2017) How do I adapt the after-effects in an aggrandized Dickey-Fuller test? [WWW] Available from https://www.quora.com/How-do-I-interpret-the-results-in-an-augmented-Dickey-Fuller-test [Accessed 23/04/18]
Jason Brownlee (2016) What Is Time Alternation Forecasting? [WWW] Available from https://machinelearningmastery.com/time-series-forecasting/ [Accessed 23/04/18]
Chris St.Jeor and Sean Ankenbruck (2018) Time Alternation for dummies- The 3 footfall action [WWW] Available from https://www.kdnuggets.com/2018/03/time-series-dummies-3-step-process.html [Accessed 22/04/18]
Pennsylvania accompaniment university (n. d) Overview of Time Alternation Characteristics [WWW] Available from https://onlinecourses.science.psu.edu/stat510/node/47 [Accessed 22/04/18]
Order a unique copy of this paper