IRJET- Artificial Algorithms Comparative Study

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 09 | Sep2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 847
ARTIFICIAL ALGORITHMS COMPARATIVE STUDY
Kaushik Indranil Patil1, Simran Sachin Sura2
1Nashik District Maratha Vidya Prasarak Samaj's Karmaveer Baburao Thakare College of Engineering, Nashik,
Maharashtra, India.
2Smt.Kashibai Navale College of Engineering, Vadgaon, Pune, Maharashtra, India.
-------------------------------------------------------------------------***------------------------------------------------------------------------
Abstract — This paper showcases a comparative study of
some commonly used Artificial Intelligence algorithms such
as Heuristics, A*algorithm, KNN, Linear Regression. In this
paper, we illustrated an example of each algorithm
describing its solution. Through this paper, the author
demonstrates the variety of problems that can be solved by
the above mentioned AI algorithms thus giving the reader
an idea about which algorithm to apply to which AI
problem.
Key Words: Artificial intelligence, Heuristics, A*, KNN,
Linear Regression.
1. INTRODUCTION
Conventional computer programming lacks the ability to
solve complex problems where the output result varies
greatly on a number of inputs Thus, It has become
necessary to develop some methods to solve a certain type
of problems which can not be efficiently solved by
conventional computer programming. As an example, let’s
consider a chess-playing program. Theoretically, there are
around legal moves possible in a chess game. Now
it is practically impossible to program a machine to handle
each and every of possible moves. Thus, it is necessary to
develop a machine which is able to play the game of chess
in a way similar to how humans play. Thus came the
evolution of Artificial Intelligence algorithms.
There doesn’t exist such an algorithm called “Unified
Artificial Intelligence Algorithm”. There are many artificial
intelligence algorithms developed but one algorithm for all
doesn't exist. Thus it becomes very necessary to be able to
recognize which algorithm to be applied to which scenario.
In this paper, we have presented five types of AI
algorithms namely Heuristics, A* algorithm, Linear
Regression, KNN.
Section I: Introduction, introduces the paper topic. Section
II: Heuristic Algorithm, discusses the analysis of heuristic
algorithm and how water jug problem can be solved using
heuristics. Section III: A* algorithm discusses the types of
pathfinding algorithms and discusses A* algorithm in-
depth with pseudo code. Section IV: Linear Regression
discusses the Linear Regression algorithm in detail,
Section V: KNN, which introduces to KNN algorithm with
some visual examples. Section VI: Concludes the paper.
2. HEURISTICS ALGORITHM
A Heuristic is a technique to solve a problem faster than
the classical method such as looping through a number of
steps until a solution is reached, or to find an approximate
solution when classic methods cannot. Heuristic
algorithms can be called a shortcut which improves
optimization, completeness, accuracy, precision and speed.
At each branching step of the heuristic algorithm, it
evaluates the information that is available and makes a
decision on which branch is to be followed next. It does so
by ranking alternatives. The Heuristic algorithm is often
effective but does not guarantee to work in every case.
2.1 Solving the Water Jug Problem
Heuristic solutions rely upon problem-solving methods
evolved from practical experience, they are rules of thumb
that try to estimate an acceptable, computationally
effective solution, as close as possible to the optimal one.
However, they do not guarantee that a solution will be
obtained at all. They are only believed to work in most
situations. There are specific heuristics that are applicable
to a particular situation, and general heuristics that are
applicable to a large range of problems. Let us consider a
heuristic for the generalized water jugs problem: if the rest
is lower than the capacity of jug A, the problem has no
solution (obvious); if the rest is a multiple of the greatest
common divisor of A and B, the problem has no solution
(the solution results from combining the differences
between A and B; e.g. with A=6 and B=3, it is impossible to
obtain a rest equal to 2, because all the differences will be
multiples of three); since the solution of this problem is
based on the water exchange between the two containers
in order to obtain the differences, there are two strategies
that lead to the solution for any problem. Let value ‘A’ and
value ‘B’ be the quantities of the two jugs, respectively. Let
rest be the quantity that is to be obtained in the first jug.
Strategy 1: while (valA != rest) { if (valB != B && valA == 0)
FillA(); if (valB == B) EmptyB(); Pour FromAtoB(); }
Strategy 2: while (valA != rest) { if (valA != A && valB == 0)
FillB(); if (valA == A) EmptyA(); PourFromBtoA(); }

Fig - 1: Efficiency of the Heuristic method. X-Axis: Steps to
Solution, Y-Axis: Efficiency.
In this way, building a solution tree is not required and
solving is done without additional memory costs.
However, since they are heuristic solutions, the adequate
strategy should be determined from case to case. For
example, if the value of A=6 and value of B=5 and the rest
is 2, the first strategy finds the solution in 6 steps, while
the latter does in 14 steps. If A=11, B=6 and the rest is 7,
the first strategy finds the solution in 24 steps, and the
latter is only 8.
3. A* ALGORITHM
A pathfinding algorithm initializes graph search by starting
at one vertex and exploring adjacent vertices until the
destination vertex is reached, generally with the intent of
finding the most optimal route but not necessarily the
most efficient. Pathfinding algorithm consists of two
phases
1. Finding the path between two given nodes.
2. Finding the most optimal path between the two nodes.
Types of pathfinding algorithm:
1. Exhaustive.
2. Eliminatory.
Let's first look at what Exhaustive algorithms are.
1. Exhaustive pathfinding algorithms
The most primitive pathfinding algorithms are Breadth-
first and Depth-first. This algorithm finds the path
between two nodes by exhausting all the possibilities,
starting from a given node. These algorithm run in O(|V| +
|E|), or linear time, where V is the number of vertices, and
E is the number of Edges between vertices.
Another yet complicated pathfinding algorithm for finding
the optimal path is the Bellman-Ford algorithm, which
yields the time complexity of O(|V||E|), or quadratic time.
However, it is not necessary to examine all the possible
paths to find the optimum one.
2. Eliminatory pathfinding algorithms
The eliminatory algorithms consist of Dijkstra and A*
pathfinding algorithms. In this project, A* algorithm is
implemented due to the high time cost of Dijkstra and the
inability of Dijkstra to evaluate negative edge weights.
3.1 A star algorithm:
A* is a variant of Dijkstra's algorithm. A* assigns a weight
to each open node which is equal to the weight of the edge
to that node in addition to the approximate distance
between that node and the finish. This approximate
distance is found by the heuristic 1 and represents a
minimum possible distance between that node and the
end. This enables it to eliminate longer paths once an initial
path is found. If there exists a path of length x between the
start and finish, and the minimum distance between a node
and the finish is greater than the distance x, that node does
not need to be examined. A* uses this heuristic to improve
on the behaviour relative to Dijkstra's algorithm. When the
heuristic evaluates to zero, A* is equivalent to Dijkstra's
algorithm[12]. As the heuristic estimate increases in
magnitude and also gets closer to the true distance, A*
continues to find optimal paths and runs faster because it
needs to examine fewer nodes than before thus less
computation required. A* examines the fewest nodes when
the value of the heuristic is exactly the true distance,
Fig - 2. Illustration of A* search algorithm[13].
A* f cost calculation equation:
f(n) = g(n) + h(n) … (1)
Where:
g is the cost of the present node from the star point

excluding the unwalkable path.
h is the cost of the present node from the ending point
including the unwalkable path.
f is the resultant cost.
Fig Fid - 3: A* pseudo code[11]
Application of A* in real-world scenarios:
1. Calculate the shortest path between two locations
on a map.
2. A* is vastly used in game development wherein it
is required to obtain the shortest path between
two points.
3. Parsing Stochastic grammars in Natural Language
Processing.
4. LINEAR REGRESSION
Let's get to know about Linear Regression algorithm with
the help of an example first, consider a situation wherein,
you are required to arrange random logs of work in
ascending order of their weight, but there is a condition:
you are not allowed to weigh each log, so you have only
one option: to guess the weight of each log just by visual
analysis. This is exactly the type of problem which can be
solved by linear regression.
The Linear regression algorithm uses the mathematical
equation: y = bx + a. This equation describes the line of
best fit for the relationship between the dependent
variable(y) and the independent variable(x).
Following equation governs the Linear regression
algorithm[16]:
Where Eq. 2 gives the y-intercept(a) and Eq. 3 gives the
slope of line(b).
Thus Linear Regression is very significant due to the
following reasons:
1. It helps understand the strength of the
relationship between the outcome and the
predictor variable.
2. Linear Regression can make an adjustment for the
effect of covariance.
3. Through Linear Regression, we can get to know
the risk factor affecting the dependent variable.
4. It also helps in quantifying new cases in our
problem statement.
Regression analysis provides us with three things[14]:
1. Description: Relation between dependent and
independent variables.
2. Estimation: Value of the dependent variable can be
estimated from observed values on the
independent variables.
3. Risk factor: Upcoming risk factor can be predicted
beforehand to avoid a critical situation in certain
analysis scenarios.
Assumptions for linear regression[15]:
1. Values of the independent variable on the X-axis is
set by the researcher.
2. The value of X-axis does not have any
experimental error.
3. All the values of Y are independent of each other
but depend only on the values on the X-axis that is
the dependent variable.

Fig - 4. A plot of cricket runs vs over, Blue Line depicts the
current score, the yellow line shows expected score.
Applications of Linear Regression algorithm:
1. Data prediction involving one dependent and one
independent variable.
2. Customer Survey analysis.
3. Market Research Studies.
5. KNN ALGORITHM
KNN is one of the most popular. KNN is used for
classification and prediction problems involving
regression. KNN is generally used in a scenario involving
multiple data points which need to be classified on the
basis of certain aspects. KNN clusters data points based on
some similarities they exhibit,
Fig - 5. Pseudo Code for KNN algorithm.
Let’s consider the example below to get a clear idea of the
algorithm:
Fig - 6. Example of KNN algorithm.
KNN is used to classify and clusterize input datasets based
on certain parameters.
Most commonly used distance functions for calculating
distance between input data points are Euclidean and
Manhattan [17][18].
= √∑ … (4)
∑ | | … (5)
Where Eq. 4 gives the Euclidean distance and Eq. 5 gives
Manhattan distance. While Euclidean distance is most
commonly used, the Manhattan method is used for
continuous variables.
Increasing the K factor makes the algorithm more
aggressive. So the classifying boundary between input data
points smoothes out. On the other hand, reducing the K
factor makes the algorithm more lenient and thus while the
classification of data points the reduced K algorithm also
considers some conflicting data points also.
Let’s have a look at the figure below showing the effect of
the K factor in the KNN algorithm[19].
Fig - 6. Classification of data points having two classified
regions with K factor kept = 1.

Fig - 7. Classification of data points having two classified
regions with K factor kept = 7.
Thus from the above figures: Fig 6 and Fig 7. We can
observe that increasing K factor makes the algorithm more
aggressive.
Keeping the K factor is up to the engineer, there is a certain
trend of value of K and the error rate produced by the
algorithm. Following is the graphical representation:
Fig - 8: Graph of Effect of K value on KNN accuracy, K value
on the X-Axis, and error rate on the Y-Axis[19].
6. CONCLUSION
In this paper, we have compared various commonly used
Artificial Intelligence algorithms with their description,
significance, mathematical model as well as their
applications. Thus we have successfully showcased the
usage of each algorithm in various real-life scenarios. From
this paper, one can differentiate between various A
algorithms and apply the most suitable algorithm on the
given problem.
REFERENCES
[1] Harris R. Creative Problem Solving: A Step by Step
Approach, Pyrczak, Los Angeles, 2002.
[2] Boldi P., Santini M., Vigna S. Measuring with Jugs,
Theoretical Computer Science, no. 282, pp. 259-270, 2002.
[3] Russell S. J., Norvig P. Artificial Intelligence: A Modern
Approach, Prentice Hall, Englewood Cliffs, New.
[4] A Heuristic for Solving the Generalized Water Jugs
Problem Bulletin of the Polytechnic Institute of Iasi, tome
LI (LV), section Automatic Control and Computer Science.
[5] Vinod Goel, Sketches of thought, MIT Press, 1995, pp.
87 and 88.
[6] Bonnie Averbach and Orin Chein, Problem Solving
Through
Recreational Mathematics, Courier Dover Publications,
2000, pp.
156.
[7] H. E. Dudeney, in Strand Magazine vol. 68 (July 1924),
pp. 97 and
214
[8] Pseudo code for Crypt arithmetic problem from
https://www.google.com/amp/s/www.geeksforgeeks.org
/solving-cryptarithmetic-puzzles-backtracking-8/amp/
[9] Desiani, Anita and Arhami, Muhammad. 2006. Konsep
Kecerdasan Buatan. Yogyakarta, Indonesia: C.V Andi Offset.
[10] Ms. Avinash Kaur, Ms. Purva Sharma, Ms.Apurva
Verma, International Journal of Scientific and Research
Publications, Volume 4, Issue 3, March 2014.
[11] A star algorithm through
https://www.youtube.com/watch?v=-L-WgKMFuhE by
Sebastian Lague.
[12] For detail A* understanding:
http://theory.stanford.edu/~amitp/GameProgramming/H
euristic
[13] A* Search Algorithm Blog, GeeksforGeeks.org
[14] Astrid Schneider, Gerhard Hommel, and Maria
Blettner; Linear Regression Analysis Part 14 of a Series on
Evaluation of Scientific Publications.
[15] Khushbu Kumari, Suniti Yadav; Linear regression
analysis study
[16] Data Science Central Article: Linear Regression:
Simple Steps and Video — Find the Equation, Coefficient
and Slope.

[17] Li-Yu Hu, Min-Wei Huang, Shih-Wen Ke, and Chih-
Fong Tsai; The distance function effect on k-nearest
neighbor classification for medical datasets.
[18] Punam Mulak, Nitin Talha; Analysis of Distance
Measures Using K-Nearest Neighbor Algorithm on KDD
Dataset.
[19] Tavish Srivastava, Article; Introduction to k-Nearest
Neighbors: A powerful Machine Learning Algorithm

IRJET- Artificial Algorithms Comparative Study

More Related Content

What's hot (18)

Similar to IRJET- Artificial Algorithms Comparative Study (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Artificial Algorithms Comparative Study