Cognitive agent system to retrieve relevant code components from a repository

Venkat Java Projects
Mobile:+91 9966499110
Visit:www.venkatjavaprojects.com Email:venkatjavaprojects@gmail.com
Cognitive Agent System to Retrieve Relevant Code Components from a
Repository
Almost all types of computer programs logics are readily available and stored
at repositories but we don’t have any facility to get required component logic
in quick simple search, almost for logic search we need to depend on Google
which will fetch required component logic with lots of noisy data and its bit
difficult for the programmer to get the required logic from that huge noisy
data.
For example if I want to search programs which implements reverse logic then
Google will give reverse logic programs with lots of noisy data. To overcome
from this issue and to ease relevant code component retrieval process we are
introducing a concept called ‘Cognitive Agent System to Obtain Relevant
Component from Repository’.
Here Cognitive Agent System is the machine training and learning process
which will be train on existing repository code component and then a train
model will be generated. Whenever user give any query (to search code
component) then Cognitive Agent System will apply train model on that new
query to get relevant code component. Always this technique will retrieve only
those components which are satisfying query and make developer work easier
to get required component logic.
To implement this project we are using following modules
Upload Code Repository: Using this module we will upload a code repository
directory which contains multiple programs for different logics. To train this
repository we will use program name as training label and program data as
train data.
Pre-process Module: Using this module we will scan each program in the
repository and remove stopwords from comments to make training size
smaller. We will remove all special symbols also.
Code2Vector: Using this module we will convert entire pre-processed data to
vector, this vector contains mapping between words and its occurrence in each

Mobile:+91 9966499110
program. In simple term vector is nothing but a matrix whose column name
contains program words and column values contains count of that words.
Query Module: Using this module we will accept query from user and then
convert query into test vector. New generated test vector will be applied on
train vector to predict all programs which contains that logic. If program
contains logic the predict function returns value > 0 and if not contains then it
will return value 0.
For example if I give query as ‘valueOf()’ which means I want to retrieve all
components which has logic for ‘valueOf()’ code. Similarly you can retrieve
code for any logic which is available in train repository.
Graph Module: Using this module we are showing graph which represents total
number of predicted components from entire train repository.
Dataset Information
To implement this project we downloaded some java repositories which
contains program from different technologies such as HADOOP, Hibernate,
apache etc. All this programs are available inside train folder.
Screen shots
To run this project double click on ‘run.bat’ file to get below screen

Mobile:+91 9966499110
In above screen click on ‘Upload Code Dataset’ button and upload train
repository
In above screen I am uploading train folder which contains more than 100
programs from different technologies. After uploading dataset will get below
screen

Mobile:+91 9966499110
Now click on ‘Preprocess Dataset’ button to read all programs and to clean all
programs by removing stop words and special symbols. Will get below screen
after preprocessing
In abovescreen wecan see it process total 101 programs and to see processed
program names then see black console. See below screen

Mobile:+91 9966499110
In above screen we can see name of each program which was processed. Now
click on ‘Convert Code2Vector’ button to convert all processed program to
vector
In above screen we can see program name and then its words and its count.
Now enter your query in text field and click on ‘Input Query & Retrieve
Component’ button to get all programs which implement that query. Once you
build model then you can search any number of queries.

Mobile:+91 9966499110
In above screen I entered query as ‘valueof’ which means I want to get all
component which used or contains logic for valueof function. Below are the
search results
In above screen we can see total 3 programs found (Amount.java,
ConcurrentReferenceHashMapTests.java and HdfsServerConstants.java)which
used or contains logic for ‘valueof’. Now I will open ‘Amount.java’ program
from train folder to see whether it really contains that valueof code or not

Mobile:+91 9966499110
In above ‘Amount.java’ from train repository we can see it contains ‘valueOf’
function. Similarly you can see any query and get component. Now click on
‘Query Result Graph’ button to get below graph
In above graph x-axis represents total components and predicted components
and y-axis represents it count.

Cognitive agent system to retrieve relevant code components from a repository

Recommended

More Related Content

What's hot (20)

Similar to Cognitive agent system to retrieve relevant code components from a repository (20)

More from Venkat Projects (20)

Recently uploaded (20)

Cognitive agent system to retrieve relevant code components from a repository