Research Work Reports The Design Computer Science Essay

Published Date: 02 Nov 2017

Keywords: Confidence Factor, Einstein Sum, Expert System, Floriculture, Fuzzy Logic, Rule Advancement Strategy.

INTRODUCTION

An Expert System is computer software that embodies a significant portion of the specialized knowledge of a human expert in a specific, narrow domain and emulates the decision making ability of a human expert [3]. Expert systems have major applicability in agriculture domain. [1] Highlight some characteristics of Expert systems in agriculture. (i) It simulates human reasoning about a problem domain, rather than simulating the domain itself. (ii) It performs reasoning over representations of human knowledge. (iii) It solves problems by heuristic or approximate methods. Expert system has been developed for identification of insect and diseases pests of varieties of crops, such as grain crop [5], vegetable crop [8], fruits [7] and economic crops [4].

FUZZY EXPERT SYSTEMS

During diagnosis sessions an expert decision depends on the user’s input feed, heuristic knowledge and basic rules for the domain which may have ambiguities and uncertainties. Fuzzy expert systems are capable of evaluating ambiguous and uncertain information using different approaches of fuzzy logic techniques.

RULE ADVANCEMENT STRATEGY USING EINSTIEN SUM

In traditional fuzzy expert systems, the confidence level once assigned to a rule during design time remains unchanged for over the years. Hence the expert system has same inference drawing ability throughout its life. For example consider

an example: IF symptom1 AND symptom 2 AND symptom 3

THEN probably disease D1 with confidence factor 0.8

For "Probably" 0.6 < CF ≤ 0.8

The CF value for this rule will remain unchanged; hence it will produce the same result with same CF value again and again over the years irrespective of user and region requirement.

To overcome this drawback a enhanced strategy namely Rule Advancement Strategy using Einstein Sum [2] is proposed. Rule Advancement Strategy mimics the core concept of rule promotion methodology which is based on the natural belief that the confidence in the rules increases gradually if the rules repeatedly result in right decision or successful sessions [6].

The Rule Advancement Strategy using Einstein Sum uses fuzzy logic approach for the calculation of Dynamic Confidence Factor (DCF). It employs conditional probability [2] of fuzzy events for the calculation of empowerment factor (EF) and Einstein sum (Fuzzy Disjunction Operator) for the computation of DCF.

The algorithm for RAS is given below:

ALGORITHM FOR RULE ADVANCEMENT STRATEGY

Let the no of times a rule fired and resulted in successful disease diagnosis is its advancement frequency denoted by (β), and let A= { RC1, RC2, RC3,…………, RCn} be the set of rule conditions that are fired for a particular disease diagnosis, and resulted in successful disease diagnosis, their existing advancement-frequency increases by one for the next diagnosis session.

Let δ1, δ2, δ3,………., δn be the existing advancement-frequency of above rule conditions, the after each session do:

Step1 IF (session is successful)

δi = δi + 1 for all x A

ELSE-IF (session is unsuccessful)

δi = δi -1 for all x A

Step2 Calculation of the empowerment factor (EF) using conditional probability as:

EF (RCi) = δi/

Here summation of all EFs’ is always nearly equal to 1.

Step3 Calculation of Dynamic Confidence Factor (DCF) using Einstein sum as:

DCF (RCi) = CF (RCi) + EF (RCi)/ [1 + CF (RCi)*EF (RCi)]

Where CF is current confidence factor, EF is empowerment factor

Using this fuzzy logic the value of DCF (RCi) always remains ≤ 1. Here DCF (RCi) = 1 indicates 100% confidence and DCF (RCi) = 0 represents 0% confidence.

COMPARISON OF PROPOSED STRATEGY WITH RULE PROMOTION METHODOLOGY [6].

The Rule Advancement Strategy algorithm is compared with rule promotion strategy suggested by [6]. The comparison is made on the datasets provided by the author [6]. The comparison results show that DCF values generated by Rule Advancement Strategy algorithm are higher than rule promotion strategy for successful diagnosis sessions. This means that the confidence level promoted by Rule Advancement Strategy algorithm after each session is higher than rule promotion methodology. While after unsuccessful diagnosis session Rule Advancement Strategy demotes the rules with greater margin. The session-wise comparisons of two approaches are given below.

SESSION ONE, STATUS: SUCCESSFUL

SESSION TWO, STATUS: SUCCESSFUL

Figure-2: Comparison of two approaches for session 2

From the above tables and figures it is clear that the DCF values by Rule Advancement Strategy are promoted higher than PCF by rule promotion methodology [6].

Session three, Status: unsuccessful

In case of unsuccessful diagnosis session the RAS algorithm demotes the fired rules with higher margin. For the provided dataset this value was 0.01. It is clear from the above tables that the confidence factor of the rules with original CF is 1 which means 100% confidence, remains unaffected irrespective of session or firing history of rule.

EXPERT SYSTEM FOR FLORICULTURE

The proposed expert system for floriculture comprises of six modules:

Knowledge Base Module.

Explanation Facility Module.

Inference Module.

User Feedback Module

Web Based GUI Module

Rule Advancement Module

IMPLEMENTATION WORK

The online fuzzy expert system for floriculture is implemented using PHP which is open source server-side scripting language. MySQL is used for creating database. The GUI made is very simple so that a user can easily interact during diagnosis. Below are some snapshots of the user- interfaces designed for expert system. These include homepage of the expert system, GUI for selecting symptoms, disease description, disease diagnosis page, description flower Rose.

Figure 4 displays the snapshot of homepage for the floriculture expert system. A provision for language selection is provided. This feature allows illiterate farmers to interact with the system in regional languages. Currently English is in operation, others are in development process

Two different login modes are also provided, administrator and expert/user. Administrator manages and controls the whole system. Expert/user login is for farmers, advisors, domain experts and other stake holders, who interacts and can give their feedback through their account.

Figure 5 shows the snapshot for the flower Rose. The basic details about the flower are given.

Figure 6 shows the snapshot of the user interface designed for the selection of symptoms. Inputs are taken through the check boxes. Finally figure 7 and 8 displays the disease diagnosed along with the CF and Pictorial conformation of disease along with solution measures.

CONCLUSION

This paper suggests an enhanced technique namely Rule Advancement Strategy using Einstein Sum to enhance the intelligence of expert system. This RAS technique is implemented in expert system of floriculture. The DCF generated by RAS technique gives marginal promotion and demotion as compared to existing rule promotion methodology. So, this approach gives a generalized expert system that changes itself according to region and user requirement. In this paper, the example of Rose flower is discussed. The system finds the exact solution with the probability of occurrence of a disease in Rose.