Job Recommendation System End To End Project || NLP || Python

hello everyone I'm back with a new another exciting video in this video I'm going to make uh tell you some new project which is based on the NLP and this is this project is a job recommendation system and it's a extension of my previous one the previous project was my job scanner in which you have to upload your resume and it will give the output of whatever the job category whether it is data science data analyst python developer uh SQL Developer and so on whatever that one so in this video it's a extension of that project and uh in it's this project is also end to endend project and uh in this project the theme is you have to give your give the name of the whatever the job post you are you want to apply and it will recommend top 10 job as per your search which is similar to your job post so those who don't know about my last uh video I will quickly show you the my last video uh which and the project so you can relate from there and uh do it your own so firstly I want to share the screen and this is my current project and here is the module all the uh models and uh the CSV files and here you can see that the CSV files look like this this is the CSV files I hope no one can understand this values Ma so firstly I want to show you my last project for that I have to go here and uh write this is my last project uh built a resume scanning web app with uh python you can easily uh understand this the concepts are too easy to understand so let me show you the uh my notebook for that I have to go there this is the notebook at the first line of code I have imported all the necessary Library which I have used here and some are in the below section then I have uh read the CSV file here and the DF is this one the raw containing the 84,000 90 and the column is 23 the rows are too much high too much High because at at the point when I was when I I was training the similarity score making the similarity score the machine got error and the error was short of storage because when the machine tried to get a similar it score whatever the row it is it will take to the RAM and function that and make a similarity score in a same point of time and uh it needs some high frequency Ram but my uh RAM storage is 2000 uh 2 GB sorry 8 GB and it needs 2 28 GB so that's why I have to remove some some rows and take some only the 10,000 rows to function my model so in the next line of code I have taken and I have seen the all the info so that I decide which column I have to take and which column I have to remove whether it is a null values missing value or duplicated value whatever it so luckily I have taken only the three column which is title and uh job description and uh the salary account I haven't used the salary column but uh I have just taken whether in the future in my coming videos maybe I will use the salary column because uh it is the extension of the next uh previous one and maybe in the future I will extend it more so in the next line of code new DF I have taken only the three column called title job description and the salary the new DF is like this it has 8490 column 90 rows and the next line of code I have taken only the 10,000 because of the space error then I have imported the analytic library and Porter stammer and stop words so that I will stop the um stop words um um then again I have imported the re random expression regular expression sorry and the next line of code I made a function uh the functions work as I will give all the rows to this function and uh it will give me the well cleaned function it's work as washing powder nma washing powder nma so uh in the clean in the clean I will give the text and uh then the If part will execute and if get any error in the If part the else will uh execute and the whatever the text uh is as it is I will take so in the first line uh the clean uh I use the r uh regular expression do Sub in which I have removed all the necessaries like uh links and add theate and so on whatever the stuff is and uh join with space in the next line of code whatever the SP whatever the line I have taken from this line of code it will go to this line and make it the lower in case and uh and the next line the word tokenization will happen and it will tokenize all the words of that truth and uh then in the uh in the last line it uh will work as it will stem the words and uh for words in clean in this clean if the word not in the stop words do word English it will remove all the stop wordss here and then uh after that after removing all all the stuff junks it will give me the it will join and returns me with space so I have used this function to my both the column title and job description then uh I have made a new column which is the addition of titles and the job description in the last column nude column basically it's uh I have operated The Joint operation here and then in the next line of code I have used tfid vectorizer and cotion similarity so tfid vectorizer uh will fit to this and it will give the Matrix and my new column so it will vector the if site is equal to one in the line and then direction is two and knowledge is three so on in similarity uh it will give the similarities course of Matrix and you can understand that the similarities work like uh uh graph 2D graph like uh uh the point in the 2D graph and the distance between the two uh point from origin and the angle between them is called similarity of that two point and the job will predicts like predicts on the base of the similarity scor so it's a num array of the similari force like this and then the brain of the project which is recommendations how it will work so the function is uh I made a function of a recommendation which gives a title the user gives a title to the recommendation and whatever the user will give the title the user give title here and whatever the title if it is equal to equal to in the title it will extract the index value it means in my sample DF if the suppose that if the user gives me uh data science so here the user gives me the data science and if the sample DF contains the data science title so extract that uh index and uh and the zeroth index of that uh data sence of that category and uh then uh after extracting the index value the let's take the rules of that index so for getting the low for getting the uh location I use the get lock of idx so that it will give me the rows of that index then I have shorted and used the list and enumerate function in the similarity course it will give to uh the similarity and uh whatever the uh rules I have taken from this it will take this and uh key is equal to Lambda uh X column X of one and the reverse is equal to true and I need only the 10 rows so after getting the 10 rows from here and then I made a null null list which is initially it's empty and uh then for I in the distance for I in distance in this row high in distance I will append all whatever the 10 values I got from here I will pend to the job in this list and uh after that I will extract uh and return to the job so here we can see that the recommendation this function here the function and the title title is administrator assist office office 10 I don't know what is the whether it's this job is exist or not or I don't know but since my uh data set is very low only the 10,000 and uh so it will give a decent output 10 output in which uh the three are same and the last two are same so uh if I will give 82,000 data it will work as fine as uh as per the user need but uh since the data set is too low and my Ram is too slow and the two short of storage that's a DEC recent project for till now and in the upcoming future I will create a new project for job job recommendation so that the accuracy will high and maybe that time the work will good or not so then after the next line of I have just remove this I have imported the pickle so that I will dump the DF files DF and then the pickle do dump the similarity is cool so this is uh this this this is as per the developer point of view whatever the what is the user point of view for that I have made a website here you can see that the website is like this one app.py I have made this website using streamlit so I have made a website only using the 31 line of code firstly I have imported the streamlet then pandas and then pickle for pickle I uh pickle I have for um loading the files I mean models then I made a recommendation DF recommendation and and uh the title The this one I have just copy pasted this one here and it will give me the job description and now come to the uh website application for website application I have uh written a title called job recommendation app and the title is equal to st. selectbox I made a select box in which the search job is here search job is available and the what what is the search job is DFA title column I will take the title column from here using DF and then job is equal to recommendation title it means job is equal to recommendation recommendation and this title we go here and job and if the job is equal to whatever the job output I will get from here the top 10 job post I will write to the uh my streamlet website so let's see in the application how it work this is the webbsite Local Host 8501 and uh here is the search box job Rec recommendation system here is written and initially it will give a job recommendation of this so let me change some clerk accountant let's search it's give me the output of similar job and uh two are similar then two are similar then four are [Music] similar okay n okay output but the since the since the data set is too low so the accuracy is not that much good as I expected so that's well and good uh I got some idea while making this project how to make a job recommendation system and uh if uh I will get the chance in the near future I will definitely make a new Rec job recommendation system which works at least better than this website so if you like this project try it your own and make and handle all the errors Maybe your uh RAM is uh much good as compared to me and your output will get more accuracy and uh try it your own and learn some new Concepts so for that I will stop here and uh hope you will enjoy you have enjoyed a lot from this project and thank you

Channel: Ashveen 002

Views: 234

Keywords: data science interview preparation, ineuron data science courses, how to crack data science interviews, data science interview book, data science interviews reddit, data science projects, ml projects resume, data science, data scientist, tech jobs, big tech, Machine Learning, Python, big data, data science at google, technical interview prep, learn to code, python, WsCube Tech, work week in my life, women in technology, work vlog, tech company

Id: LWzlEn7e8p8

Length: 22min 50sec (1370 seconds)

Published: Wed Feb 14 2024