Skip to Main Content
 

Global Search Box

 
 
 
 

Files

ETD Abstract Container

Abstract Header

BALLWORLD: A FRAMEWORK FOR LEARNING STATISTICAL INFERENCE AND STREAM PROCESSING

Abstract Details

2017, Master of Science in Engineering, University of Akron, Computer Engineering.
This thesis presents a framework called \emph{BallWorld} that enables the exploration and understanding of topics in the areas of statistical inference and data science. Many fundamental concepts in these areas are abstract and difficult to reason about using examples that can be carried out by hand. BallWorld offers users a concrete paradigm in which such abstract concepts can be represented as attributes of a physical system. The concrete paradigm also allows users to implement techniques to extract value from available data. The results can also be interpreted in the same concrete setting. BallWorld is a window into a collection of streams; each stream is a sequence of balls with different attributes such as shapes, sizes or colors. Balls enter the windows from the left end leave from the right end. The size of the window is configurable. Inference algorithms can only work with the balls (data) that have passed through the window at any given time. Balls that have not yet entered the window are unknown and balls that have exited the window cannot be retrieved. One can design and implement algorithms and techniques drawn either from statistical inference or from data science to make inferences about the streams of data in BallWorld. The core framework for BallWorld has been designed and implemented. As an initial step, three well-known methods of statistical inference, namely the Method of Moments, Maximum Likelihood Estimation and Hypothesis Testing have been implemented. In addition, three stream processing algorithms, namely Counting the Number of Distinct Elements, Estimating Moments and Filtering have been implemented. In order to test and validate BallWorld an injection module was designed and implemented to inject balls with attributes drawn from a variety of well-known probability distributions. Experimental results demonstrate how inferences made using the statistical inference techniques and the stream processing algorithms are related. In the future, additional techniques and algorithms can be implemented to visualize, archive and make inferences from streams of data for a variety of applications.
Shiva Sastry (Advisor)

Recommended Citations

Citations

  • Ravali, Y. (2017). BALLWORLD: A FRAMEWORK FOR LEARNING STATISTICAL INFERENCE AND STREAM PROCESSING [Master's thesis, University of Akron]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=akron1498769835817335

    APA Style (7th edition)

  • Ravali, Yeluri. BALLWORLD: A FRAMEWORK FOR LEARNING STATISTICAL INFERENCE AND STREAM PROCESSING. 2017. University of Akron, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=akron1498769835817335.

    MLA Style (8th edition)

  • Ravali, Yeluri. "BALLWORLD: A FRAMEWORK FOR LEARNING STATISTICAL INFERENCE AND STREAM PROCESSING." Master's thesis, University of Akron, 2017. http://rave.ohiolink.edu/etdc/view?acc_num=akron1498769835817335

    Chicago Manual of Style (17th edition)