Searching String in Big-Data: A Better Approach by Applied Machine Learning

Abstract

There are hot patterns of huge data sets in numerous regions in the course of the most recent 5 years. Looking through string or pattern from a gigantic record is intense generally in the event of randomized situation. These require the advancement of new algorithmic strategies to investigate such enormous data sets and solve optimization errands utilizing sorting on indexing levels and Applied Machine Learning Models. One of the essential strides in text processing is string searching and pattern matching. A word search algorithm works by finding the first or all the occurrences of a word in a textual data or ASCII files. The pre-processing phase is used to determine the formula for number of positions by which the pattern needs to be shifted in case of a mismatch in the matching phase. The fundamental objective of string search or pattern matching algorithms is to increase efficiency by reducing the number of comparisons and increase the length of shifts in event of a mismatch. The issue of efficiency of string search algorithms has probably never been considered so seriously and genuinely until the virtual content explosion caused by the web and the task of mining valuable data and information from it. In this paper, a better search algorithm “Tara–Paras String Search” is introduced that is faster than conventional Binary Search and Interpolation Search. Indexing levels are introduced by length of the word, sequence total of alphabets and starting letter of the word to reduce the size of input. For analysis, 2 data sets are considered. The dictionary of English words having more than 109,000 words and a list of more than 2.5 Lac sorted numbers and uniformly distributed (multiples of 7) are taken for data sets. Analysis and Implementations Models have been implemented, compared and executed in Python with time complexity and obviously Applied Machine Learning will select the faster one.