로보틱스 업무 자동화(RPA)
DEEP-LEARNING DECISION MAKING with RPA
2019.08.05.
2
2
Question for Deep-Learning Decision Making RPA
Discover Black Box Process (“Unknown Business Logic”)
Given Features
Process
RPA
Human Decision
Decision
Flow
ON BOT PC
BLACK BOX PROCESS
NOT (Something) + (Something) = 5
10 (Somethig) 2 = 5
PROCESS AUTOMATION NEEDS DEEP DECISION MAKING
RATHER THAN DEEP COGNITIVE TECH
3
3
Realization Steps for Deep-Learning Decision Making RPA
Configure Features In RPA Environment
- Features should be found in PC
X1 = News Cast ID (“NC_{Digits}”)
X2 = Relative Page Number
<Known Features>
Y = 0 OR 1
<Final Decision>
<Unknown Business Logic>
BLACK BOX PROCESS
<ANSWER For Unknown Business Logic>
IF (Digits from X1 can be divided by X2) AND (X2 is not 1), THEN Y = 1
ELSE Y = 0
4
4
Realization Steps for Deep-Learning Decision Making RPA
Crawling Train Dataset For DDM Model
* 1,000 Times Refresh
Extracting Train Data
No preprocess in crawling
5
5
Realization Steps for Deep-Learning Decision Making RPA
Deep-Learning Decision Making RPA – REQ/RES PROCESS
BOT
Gain Features
Request Data DDM Model
Handing
BLACK BOX PROCESS
Making Decision
Handing
Response Data
BOT
DO
Decision
Flow
Trained
6
6
Realization Steps for Deep-Learning Decision Making RPA
Deep-Learning Decision Making RPA – REQ/RES PROCESS
Model
Request Data Response Data
BOT Activity Per Decision
7
7
Realization Steps for Deep-Learning Decision Making RPA
How to run Source Code
DDM_Model.py
# Requirements
1. Numpy – pip install numpy
2. Pandas – pip install pandas
3. Sklearn – pip install sklearn
4. Keras (tensorflow) – pip install tensorflow; pip install keras
“DEEP_WORK.zipDEEP_WORKDEEP_DECISION_MAKING_RPADEEP_DECISION_MAKING.xaml”
MAIN XAML FILE

Deep learning decision making with rpa

  • 1.
    로보틱스 업무 자동화(RPA) DEEP-LEARNINGDECISION MAKING with RPA 2019.08.05.
  • 2.
    2 2 Question for Deep-LearningDecision Making RPA Discover Black Box Process (“Unknown Business Logic”) Given Features Process RPA Human Decision Decision Flow ON BOT PC BLACK BOX PROCESS NOT (Something) + (Something) = 5 10 (Somethig) 2 = 5 PROCESS AUTOMATION NEEDS DEEP DECISION MAKING RATHER THAN DEEP COGNITIVE TECH
  • 3.
    3 3 Realization Steps forDeep-Learning Decision Making RPA Configure Features In RPA Environment - Features should be found in PC X1 = News Cast ID (“NC_{Digits}”) X2 = Relative Page Number <Known Features> Y = 0 OR 1 <Final Decision> <Unknown Business Logic> BLACK BOX PROCESS <ANSWER For Unknown Business Logic> IF (Digits from X1 can be divided by X2) AND (X2 is not 1), THEN Y = 1 ELSE Y = 0
  • 4.
    4 4 Realization Steps forDeep-Learning Decision Making RPA Crawling Train Dataset For DDM Model * 1,000 Times Refresh Extracting Train Data No preprocess in crawling
  • 5.
    5 5 Realization Steps forDeep-Learning Decision Making RPA Deep-Learning Decision Making RPA – REQ/RES PROCESS BOT Gain Features Request Data DDM Model Handing BLACK BOX PROCESS Making Decision Handing Response Data BOT DO Decision Flow Trained
  • 6.
    6 6 Realization Steps forDeep-Learning Decision Making RPA Deep-Learning Decision Making RPA – REQ/RES PROCESS Model Request Data Response Data BOT Activity Per Decision
  • 7.
    7 7 Realization Steps forDeep-Learning Decision Making RPA How to run Source Code DDM_Model.py # Requirements 1. Numpy – pip install numpy 2. Pandas – pip install pandas 3. Sklearn – pip install sklearn 4. Keras (tensorflow) – pip install tensorflow; pip install keras “DEEP_WORK.zipDEEP_WORKDEEP_DECISION_MAKING_RPADEEP_DECISION_MAKING.xaml” MAIN XAML FILE