Learner Performance-based Behavior Optimization Algorithm: A Functional Case Study

. A novel algorithm called learner performance-based behavior algo-rithm (LPB) was proposed for single and multi-objective by Chnoor M. Rahman and Tarik A. Rashid in 2021. LPB proved its ability to deal with complex optimization problems compared to the dragonfly algorithm (DA), genetic algorithm (GA), and particle swarm optimization ( PSO ). This paper presents and explains the implementation of the LPB algorithm, and it applies it as a model in a case study to maximize a fitness function. As a result, the LPB algorithm is successfully improved the initial population and achieved the optimal solution.


Introduction
Metaheuristic refers to higher-level heuristics, which have been developed for solving a variety of optimization problems. Various metaheuristic algorithms have recently been successful in tackling intractable situations. The advantage of employing these algorithms to solve complicated problems is that they produce the approximate solutions in a short amount of time, even for very complex problems [1]. Optimization is used in almost all areas of our lives, such as engineering, medicine, business planning, control, energy, etc. These algorithms intelligently select the best solutions from a wide range of choices [2]. Some widely used optimization algorithms are extracted from natural systems, such as Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO) [3].
The rest of the paper is organized into sections, where section two explains LPB in brief. In section three, a case study has been designed to evaluate LPB. Finally, in section four, the main points are concluded.

Learner Performance-based Behavior Algorithm
LPB algorithm is inspired by the idea of accepting graduates from high school to university. Through the steps that are applied during the admission of Learners, which are the methods used to divide learners and group them according to their cumulative rate. Also, these methods are used to improve the behavior and level of performance of the individuals after admission to the departments. Learners need to use new study habits because the methods they used to study in junior high do not work properly in college [6,7,8]. In this algorithm as a first step, a group of individuals is selected from the population. These individuals are then divided into sub-groups, and the best individuals are then selected from the subgroups depending on their fitness. Their behavior and performance are then improved by having them work as groups. Where teamwork will provide information sharing among themselves when they study (crossover), this method will affect their behavior randomly (mutation). LPB uses crossover and mutation techniques of the genetic algorithm. LPB algorithm works as shown in Figure 1.
Step 1: let M consists of 16 individuals. Then evaluate the fitness of all M individuals by the fitness equation. Calculate the summation, average and find maximum fitness form M as shown in Table 1:  Table 2. Step 3: Compare M individuals with O Good and Bad highest fitness to divide M. Step 4: Check if the Perfect population is not empty, select individuals from PF. If the PF is empty select individuals from the Good population (GP) when it's not empty, if the GP is empty, then select individuals from the Bad population (BP). The selected individuals will be used in the crossover operator. Note that the number of selected individuals equals the number of required individuals N, which we specify in the first step. Step 5: Apply crossover between selected individuals from Table 4 and good individuals from Table 2. Step 6: Apply mutation on new individuals (Child), to maximize the function (randomly convert 0 --> 1) 1bit for each individual. New individuals are shown in Table 7: Step 7: Calculate the fitness of new individuals as shown in Table 8. Step 8: Find the sum, average, and max between new individuals from Table 8 and the parents (Pi) from Table 5.  The previous steps (2 to 8) will be repeated till the required number of iterations or the stop condition is met, then the optimal solution is returned.
Repeat steps 2 to 8, to do the second iteration.
Step 2: Create a new O sub-population from new M population table 10.

Bad
Step 4: Check if the Perfect population is not empty select individuals from PF, if the PF is empty, select individuals from the good population when it's not empty, if the GP is empty, select individuals from the BP. The selected individuals will be used in the crossover operator.
Note: The number of selected individuals will equal the number of required individuals N, which we specify in the first step.

Bad
Step 5: Apply crossover between selected individuals from Table 13 and good individuals from Table 11. Step 6: Apply mutation on new individuals (Child) to maximize the function (randomly convert 0 bit to 1 bit) 1bit for each individual. New individuals are shown in Table 16. Step 7: Calculate the fitness of new individuals by fitness equation, as shown in Table  17.  Thus, it can be noted from the results that the population has been improved and the efficiency of individuals is increased. Table 19 shows a comparison between the summation and the average and the optimal value for each iteration.

Conclusion
This study presents Learner based Performance algorithm. A case study is designed to describe the crossover and mutation processes that may confuse readers of this algorithm. In the experimental results, LBP proved its potential in improving and developing qualities and likewise finding the optimal solution. LPB improves and obtains better solutions iteratively. The authors recommend that this algorithm can be enhanced to reduce the processing time. In addition, changing the GA's crossover and mutation operations with other operations or mathematical equations from other competitive optimization algoirhtms might lead to a better performance of LBP.