Re: JESS: Storing rules in an excel sheet

Hello Wolfgang,

it is kind of related to your scenario, but our data mining does return
some rules as well.
Here are some examples we "mine":

Linear regression:

- 0.216 * B_SZ_1 + 0.685 * L_SZ_1 + 0.392 * R_SZ_2 + 0.867

polynomial regression:

0.000 * BETA_SZ ^ 4.000 + 0.035 * B_SZ_1 ^ 1.000 - 0.103 * L_SZ_1 ^
3.000 - 0.163 * R_SZ_2 ^ 5.000 - 0.002 * R_SZ_3 ^ 5.000 - 0.126

W-REPTree:

L_SZ_1 < 2.75

| B_SZ_1 < 2.75

| | R_SZ_2 < 0.45

| | | R_SZ_3 < 0.45

| | | | B_SZ_1 < 2.25 : 2.38 (4/0) [2/0]

| | | | B_SZ_1 >= 2.25 : 2.1 (2/0.01) [4/0.02]

| | | R_SZ_3 >= 0.45 : 2.14 (4/0.01) [1/0.04]

| | R_SZ_2 >= 0.45

| | | B_SZ_1 < 2.25 : 2.49 (11/0.04) [5/0.03]

| | | B_SZ_1 >= 2.25 : 2.33 (10/0.02) [8/0.02]

| B_SZ_1 >= 2.75

| | R_SZ_2 < 0.75

| | | BETA_SZ < 57.5

| | | | R_SZ_2 < 0.45 : 1.99 (5/0.01) [1/0]

| | | | R_SZ_2 >= 0.45 : 2.08 (3/0) [3/0]

| | | BETA_SZ >= 57.5 : 2.11 (2/0) [1/0]

| | R_SZ_2 >= 0.75 : 2.25 (5/0.01) [3/0.05]

L_SZ_1 >= 2.75

| R_SZ_2 < 0.45

| | R_SZ_3 < 0.45

| | | BETA_SZ < 57.5 : 2.43 (8/0.03) [4/0.04]

| | | BETA_SZ >= 57.5 : 2.59 (4/0.01) [2/0.06]

| | R_SZ_3 >= 0.45

| | | BETA_SZ < 53.75 : 2.59 (3/0.02) [0/0]

| | | BETA_SZ >= 53.75 : 2.53 (3/0.01) [3/0.02]

| R_SZ_2 >= 0.45

| | B_SZ_1 < 2.25

| | | BETA_SZ < 53.75

| | | | R_SZ_2 < 0.75 : 2.56 (2/0.01) [1/0]

| | | | R_SZ_2 >= 0.75 : 2.83 (3/0.01) [0/0]

| | | BETA_SZ >= 53.75 : 2.8 (5/0) [4/0.07]

| | B_SZ_1 >= 2.25 : 2.62 (26/0.03) [9/0.01]


W M5Rules:

Rule: 1

IF

L_SZ_1 > 2.75

R_SZ_2 > 0.45

THEN

TEPS = + 2.617 [50/66.193%]

Rule: 2

IF

L_SZ_1 > 2.75

THEN

TEPS = + 2.4476 [27/74.978%]

Rule: 3

IF

B_SZ_1 <= 2.25

THEN

TEPS = + 2.377 [24/88.231%]

Rule: 4

IF

R_SZ_2 <= 0.75

B_SZ_1 <= 2.75

THEN

TEPS = + 2.1759 [18/80.216%]

Rule: 5

IF

R_SZ_2 > 0.75

THEN

TEPS = + 2.2709 [17/80.919%]

Rule: 6

TEPS = + 2.0447 [15/100%]


I hope it will help to understand. Our intention is to evaluate a model
designed in catia (get all parameters via xml or something else) and use
our rulebase to verify if it is possible to manufacture.
Florian

> I'm intrigued. Would this be, in any way, related to this scenario:
> We have observations of a, b from C = set of complex numbers, and the
> resulting conic section (circle, ellipse,...): how to derive the type
> of the latter from the properties of the former?
>
> Then, you aren't storing rules; you would like to derive rules from
> the observed data?!
>
> -W
>
> On 23 February 2011 18:21, Flogger<[email protected]>  wrote:
>> Thank you,
>>
>> let me try to explain what we are planning to do and why our first thought
>> was using an excel sheet.
>> We are doing research on sheet metal forming and my tutor is doing some
>> simulation and using rapidminer to create rules explaining the forming.
>> These rules are formulas expressing the natural strain. He wants to store
>> these rules in a multidimensional database (many sheets in an excel
>> workbook). One sheet stands for one geometrical object.
>> I know it is not the best idea but we do not know what would be the best way
>> of storing our rules.
>> It is important that the rules are not stored directly into jess code,
>> because my tutor wants to autogenerate his new rules without working with
>> jess.
>>
>> Thank you again
>> Florian
>>
>> Am 23.02.2011 17:50, schrieb Friedman-Hill, Ernest:
>>> Jess rules are textual, like program code. Typically what you're going to
>>> store in a spreadsheet is a decision table, which some simple rule engines
>>> will execute directly. That isn't what Jess does. You could write a Jess
>>> program to *implement* a decision table runner, or you could try to store
>>> Jess rules as text in spreadsheet cells (although I have no idea why anyone
>>> would ever want to do that!)
>>>
>>>> -----Original Message-----
>>>> From: [email protected]
>>>> [mailto:[email protected]] On Behalf Of Flogger
>>>> Sent: Wednesday, February 23, 2011 11:13 AM
>>>> To: jess-users
>>>> Subject: JESS: Storing rules in an excel sheet
>>>>
>>>> Hi Folks,
>>>>
>>>> first of all I want to introduce myself.
>>>> My name is Florian. I am from Germany. I am 22 years old and I am
>>>> writing my bachelor thesis about expert systems.
>>>> My tutor wants me to store my rule database in excel because
>>>> he wants to
>>>> automatically generate it using a data mining software.
>>>> My question right now is: Is it possible? And if yes, how?
>>>> I read "Jess in action" but there was only an explanation for
>>>> storing in
>>>> xml files.
>>>>
>>>> Thank you.
>>>> Florian
>>>>
>>>>
>>>> --------------------------------------------------------------------
>>>> To unsubscribe, send the words 'unsubscribe jess-users
>>>> [email protected]'
>>>> in the BODY of a message to [email protected], NOT to the list
>>>> (use your own address!) List problems? Notify
>>>> [email protected].
>>>> --------------------------------------------------------------------
>>>>
>>>
>>> --------------------------------------------------------------------
>>> To unsubscribe, send the words 'unsubscribe jess-users [email protected]'
>>> in the BODY of a message to [email protected], NOT to the list
>>> (use your own address!) List problems? Notify [email protected].
>>> --------------------------------------------------------------------
>>>
>>>
>>
>>
>> --------------------------------------------------------------------
>> To unsubscribe, send the words 'unsubscribe jess-users [email protected]'
>> in the BODY of a message to [email protected], NOT to the list
>> (use your own address!) List problems? Notify [email protected].
>> --------------------------------------------------------------------
>>
>>
>
> --------------------------------------------------------------------
> To unsubscribe, send the words 'unsubscribe jess-users [email protected]'
> in the BODY of a message to [email protected], NOT to the list
> (use your own address!) List problems? Notify [email protected].
> --------------------------------------------------------------------
>
>