acn_ht_en
| Topic List |
| Cyber Analytics | | JSDataSet a JavaScript Dataset statistical free library | | Time series vertical analytics |
| Forums List |
| Scrum: Why What How | | Breakdown Structure Method (BSM) | | Insights my php mysql websites | | Real-Time Big-Data Analytics Technologies | | Information systems design |
|
Time series vertical analytics |
 Gideon Koch
|
| Real-time big-data analytics time series vertical algorithm parallel multithreading multiprogramming large grid processing distributed document-based database.
Time series is vertical analytics is used to analyze, open structure (JSON) time evolvement events, and predict the future whether it is a previous event or new behavior.
Horizontal analytics uses Machine learning as a forecasting algorithm.
The data is stored in ONE record with many fields.
The result is a forecast based on defined fields and content.
The information is streaming in high velocity and in variety, nearly constant rate of reporting.
The analytics is done on the vertical evolvement of the data as time passed.
The process to convert, vertical data into horizontal information is called, pattern pivot transformation.
This process creates a Cumulative Analysis Pattern Pack (CAPP)
Analyze and forecasting information, is streaming powerfully every few milliseconds.
This requires a special technique to provide both historical and fast (Lambda/CRISP-DM) to provide required forecasting on time.
Popular vertical algorithms:
+ Autoregression (AR)
+ Moving Average (MA)
+ Autoregressive Moving Average (ARMA)
+ Autoregressive Integrated Moving Average (ARIMA)
+ Seasonal Autoregressive Integrated Moving-Average (SARIMA)
+ Seasonal Autoregressive Integrated Moving-Average with Exogenous Regressors (SARIMAX)
+ Vector Autoregression (VAR)
+ Vector Autoregression Moving-Average (VARMA)
+ Vector Autoregression Moving-Average with Exogenous Regressors (VARMAX)
+ Simple Exponential Smoothing (SES)
+ Holt Winter’s Exponential Smoothing (HWES)
+ Prophet
In most cases, this type of analytics involves a real-time process on, NoSQL database using Lambda Architecture and CRISP-DM methodology. It also requires in-depth knowledge of distributed data technologies such as parallel processing, document (JSON) based database, and much more.
This forum provides insights into the algorithms and in-depth technologies such as SQL/NoSQL, pivot, and multithreading/multiprocessing using python as well as PL/SQL, partitions, and many more technologies. |
|
|
| Horizontal vs Vertical Analytics |
Gideon Koch
|
| While horizontal analytics base on Machin Learning (ML) uses one record (line) with many fields, vertical analytics uses many records (Big-Data) streaming in high velocity in peak capacity, usually at a fixed time rating.
The difference in analyzing these two types of data is huge.
Horizontal analytics is usually static information gathered once and analyzed with no time requirements.
The information is usually stored using a Python dataset in pandas which loads the data once and provides an answer.
Vertical analytics gather information fast and required to provide speed and batch processing under time constraints.
Under Bug-Data streaming required to provide answers fast, we need some NoSQL technologies in order to cope with the requirement such as Lambda architecture.
The main research in such algorithms are:
> Timing (Volume and Velocity) parallel, NoSQL stored process
> Horizontal unknown extensions in the incoming information (Variety)
> Stored historical periods for prediction purposes (SARIMA)
A new topic investigated is the prediction of new events (Veracity).
We would like that the algorithm will process new external factors that have not yet been measured.
The study focuses on determining external events such as new observations and their impact on the prediction formula.
To provide an accurate prediction, we search for additional algorithms.
The impact factor algorithm defines the impact of the new streaming data on the main algorithm as input parameters.
|
|
|
|