A Knowledge Based System (KBS) using CLIPS for Horticulture disease and pest diagnosis that automatically generates rules from a .txt file. Read more about the automation

View the data about the symptoms from the symptoms documentation

Clone the repo:

git clone https://github.com/wcyn/clips-horticulture-expert-system

Run the CLIPS Program:

 Click on File -> Load, then browse to the directory with the cloned files.
 Click on diagnosis_rules_automated.CLP file. Then click Open.

On the CLIPS command line, type:

CLIPS> (run) 

How the Weights Work
Template Definitions
 deftemplate symptom-details
 deftemplate disease-weight
Function Definitions
 deffunction ask-question
 deffunction yes-or-no-p
 deffunction which-plant
 deffunction diagnose-plant
Query Rules
 defrule determine-plant
Automated Query Rules for Specific Symptoms
 deffunction read-from-symptoms-file
 deffunction create-query-rules
  defrule determine-yellow-patch-leaves
Diagnosis Rules
 deffunction create-check-diagnosis-rule
 deffunction create-diagnosis-rules
Giving Advice
 deffunction give-advice

Each symptom has an assigned certainty factor (a number between 0 and 1), which defines the confidence level of the prescence of the disease if the symptom is true. The certainty factors are merely opinions and are therefore not supported by statistical evidence.

For example, Rose Rust has four symptoms:

• yellow-patch-leaves (certainty factor: 0.4)
• orange-spores-leaves (certainty factor: 0.5)
• leaves-fall (certainty factor: 0.3)
• plants-defoliated (certainty factor: 0.6)

As such:

• yellow-patch-leaves => rose-rust (by 0.4)
• orange-spores-leaves => rose-rust (by 0.5)
• leaves-fall => rose-rust (by 0.3)
• plants-defoliated => rose-rust (by 0.6)

If yellow-patch-leaves and leaves-fall are set to true. Then we shall obtain a confidence level of the rose flower having rose-rust through a number of calculations:

• yellow-patch-leaves => not(rose-rust) by a factor of 0.6 and leaves-fall => not(rose-rust) by a factor of 0.7
• yellow-patch-leaves + leaves-fall => not(rose-rust) by a factor of 0.6 * 0.7 = 0.42
• Which, after inverting gives us: yellow-patch-leaves + leaves-fall => rose-rust by a factor of 1 - 0.42 = 0.58

If we set the threshold to 0.7, the diagnosis for rose-rust would fail.

But if orange-spores-leaves symptom was also present, we would have a confidence factor of (1 - (0.6 * 0.7 * 0.5)) = 1 - 0.21 = 0.79, and since 0.79 > 0.7, the diagnosis for rose-rust would be true.

These provide a framework to hold the various groups of data items.

Description: Defines the types of data describing the details of a symptom
Slots:
-- symptom-name - Name of the Symptom
-- plant-name - The name of the plant
-- disease-or-pest - What disease or pest the symptoms belongs to
-- prescence - If the symptom is present in the plant or not. Can be yes or no. Default is no
-- weight - The weight that the symptom contributes to the overal disease or pest

Description: Defines the name of the disease and the total weight it has after adding up all its symptom's weights
Slots:
-- disease-or-pest-name - Name of the disease or pest
-- plant - Name of the plant that the disease belongs to
-- weight - The total weight that the symptoms of the disease contribute. It's an addition of only the present symptoms.

Description: Asks a question
Arguments:
-- question - The question to ask the user
-- allowed-values - Input values accepted for the question asked
Steps:
-- Printout the question
-- Get user input and store it in answer
-- If answer is not in allowed-values, keep asking the question
-- Finally, return answer

Description: Asks question and gets yes or no response from user
Arguments:
-- question - The question to ask the user
Steps:
-- Get yes or no response from user and store that in response
-- Return yes if response is yes and no if response is no

Description: Sets analysis mode to true or false. Analysis mode allows the user to see the calculations that lead to the diagnosis
Arguments:
-- question - Response by the user, yes or no
Steps:
-- Get response from user and change analysis gloabal variable to TRUE or FALSE accordingly.

Description: Finds what plant is affected
Arguments:
-- question - The question to ask the user
Steps:
-- Get number / identity of plant affected from the user and store that in response
-- Return cabbage if response is 1, banana if 2 and so on..

Description: Get weight totals of the group of symptoms for a disease or pest and give a positive diagnosis if threshold is exceeded
Arguments:
-- plant-name - The name of the plant
-- disease-or-pest - What disease or pest the symptoms belongs to
-- threshold - The total weight that must be exceeded for a certain disease or pest to be the accepted diagnosis
Steps:
-- Initialize the weight to zero
-- Get all the weights of facts in the symptoms-details template whose prescence is yes, plant-name is [argument specified for plant-name] and whose disease-or-pest is [argument specified for disease-or-pest]
-- Add up those weights and assign them to weight b -- If weight is greater than [argument given for threshold], then return TRUE

Description: Determines the type of plant affected
Rule conditions: Only fires if no diagnosis has been reached, and no plant-name has been provided by the user

Rules used to query the user about the various symptoms are automatically generated from the symptoms.txt file. For all the rules to be generated, the function [read-from-symptoms-file](#deffunction read-from-symptoms-file) is called, which loops through the file extracting the data that is needed to create the rules.

Description: Loop through the text file with the data and use the create-query-rules function to create rules for each symptom
Arguments:
-- template - The template to use for the details of the symptoms
-- file - The name of the file containing the data

(deffunction read-from-symptoms-file (?template ?file)
    (open ?file file-data) ; open the file and store data in file-data
    (bind ?stop FALSE) ; initialize stop variable to FALSE
    (bind ?plant-name (read file-data)) ; 1st line of the beginning of a new pest or disease is the plant name
    (bind ?disease-or-pest (read file-data)) ; 2nd line of the beginning of a new pest or disease is the disease or pest name
    (while (not ?stop) ; while stop variable is not TRUE
        (bind ?temp-line (readline file-data)) ; read entire line from text file
        (if (eq ?temp-line EOF) ; if End of File
            then (bind ?stop TRUE) ; Set stop variable to TRUE
        else (if (eq ?temp-line "ENDGROUP") ; If "ENDOFGROUP" check for the diagnosis of the disease or pest
            then
            (create-check-diagnosis-rule ?plant-name ?disease-or-pest)
            (bind ?plant-name (read file-data)) ; Read plant name of the group of symptoms
            (bind ?disease-or-pest (read file-data)) ; Read disease or pest name of the next group of symptoms
        else (if (eq ?temp-line "") ; If reads empty string, do nothing
                then (printout t "") ; Do nothing
        else
            (bind ?exp-line (explode$ ?temp-line)) ; delimit the line read using spaces
            (create-query-rules ;create the rules needed to query the user
                ?template 
                ?plant-name
                ?disease-or-pest
                (implode$ (subseq$ ?exp-line 1 1))
                (implode$ (subseq$ ?exp-line 2 2))
                (implode$ (subseq$ ?exp-line 3 3)))
            ))))
    (close)) ;close the file when done

Description: Generate a rule to query the user about a symptom
Arguments:
-- template - The template to use for the details of the symptoms
-- plant-name - The name of the plant
-- disease-or-pest - What disease or pest the symptoms belongs to
-- symptom - The name of the symptom to ask the user about
-- qn - The question to ask the user
-- weight - The weight that the symptom contributes to the overal disease or pest

The function code is as shown:

(deffunction create-query-rules (?template ?plant-name ?disease-or-pest ?symptom ?qn ?weight)
    (bind ?symptom-rule-name (str-cat "determine-" ?symptom))
    (build (str-cat
            "(defrule " ?symptom-rule-name
                "(not (diagnosis ?))
                 (plant-name " ?plant-name ")
                =>
                (assert
                    (" ?template 
                        "(symptom-name " ?symptom ")
                        (plant-name " ?plant-name ")
                        (disease-or-pest " ?disease-or-pest ")
                        (prescence 
                            (yes-or-no-p " ?qn "))
                        (weight " ?weight "))))"
            )))

If this function is run as such (remember that it depends on the yes-or-no-p function):

CLIPS> (create-query-rules symptom-details rose rose-rust yellow-patch-leaves "\"Yellow patched leaves? \"" 20)

It will generate a rule called determine-yellow-patch-leaves that looks like this:

Description: Determines whether the rose plant has yellow patches on its leaves
Rule conditions: Only fires if no diagnosis has been reached and plant-name is rose
If these conditions are reached, assert the symptom-details facts. Also ,ask the user a yes / no question and assign the answer to presence.

(defrule determine-yellow-patch-leaves ""
   (not (diagnosis ?))
   (plant-name rose)
   =>
    (assert
        (symptom-details 
            (symptom-name yellow-patch-leaves)
            (plant-name rose)
            (disease-or-pest rose-rust)
            (prescence 
                (yes-or-no-p "Yellow patched leaves? "))
            (weight 25))))

Description: Generate a rule that prompts the diagnosis to be checked after certain questions have been answered
Arguments:
-- plant-name - The name of the plant
-- disease-or-pest - What disease or pest the symptoms belongs to

If this function is run like so:

CLIPS> (create-check-diagnosis-rule rose rose-rust)

It will generate a rule called check-rose-rust-diagnosis that looks like this:

(defrule check-rose-rust-diagnosis
   (not (diagnosis ?))
   (plant-name rose)
   =>
   (assert (check-rose-rust-diagnosis)))

Description: Generate a rule to confirm whether a plant has a certain disease / pest or not.
Arguments:
-- template - The template to use for the details of the symptoms
-- plant-name - The name of the plant
-- disease-or-pest - What disease or pest the symptoms belongs to
-- diagnosis-st - What to print out to the user if the Diagnosis is positive

If this function is run like so:

CLIPS> (create-diagnosis-rules rose rose-rust "\"Your Rose seems to be suffering from Rose Rust\"")

It will generate a rule called confirm-rose-rust that looks like this:

(defrule confirm-rose-rust
   ?f <- (check-rose-rust-diagnosis)
   =>
   (retract ?f)
   (if (diagnose-plant rose rose-rust 0.7)
      then
      (assert (diagnosis "Your Rose seems to be suffering from Rose Rust" rose rose-rust))))

After a diagnosis has been given, the system offers some advice to the user depending on the disease or pest that the plant has been diagnosed with. For example, it may suggest a link to a website with helpful information.

The function that takes care of this is called give-advice

Description: Get advice from a text file depending on the pest or disease and display it to the user
Arguments:
-- plant-name - The name of the plant
-- disease-or-pest - What disease or pest the symptoms belongs to
-- filename - The filename where the advice data is stored
Steps:
-- Get the plant name and disease or pest name from the file -- Check if that matches with the given plant name and disease or pest name
-- If they match, print out the advice that follows until you reach the text ENDGROUP
-- If not, keep reading through the file