howardvickers / resume-match

Web app to match resume to job type, using nlp svm classifier model. Data via webscraping. Uploaded resume converted from PDF to text using OCR.

resume-match.net

Project Introduction

This is a personal project to create a web app that allows someone with no coding experience to run machine learning algorithms on business data.

Machine Learning and Data Science present two main barriers to entry:

• Coding skills (eg python, R, etc)
• Theoretical background (eg statistics, math etc)

This web app explores how the first barrier can be overcome with a point-n-click interface to use machine learning techniques that normally require good knowledge of python, pandas, numpy and similar programming skills.

The second barrier could be addressed with step-by-step guidance and training; this is considered beyond the immediate scope of this project.

Note that the web app is intended as a conceptual project rather than a commercial product. It aims to showcase the programmer's combined skillset of data science and web dev.

Business Context

This project is guided by a sense of a hyperthetical, yet practical business situation: a startup has sales or other data that is available as a csv file and wishes to gain insights into the data; perhaps a desire to predict future sales and to understand where to employ resources for the best performance.

Functionality

The web app offers the following functionality:

• upload csv files
• view initial rows (equivalent to df.head() in pandas)
• change column names
• select columns for regression analysis
• select columns using jquery (without page refresh)
• compare regressions with different tools (linear, random forest, etc)
• demo mode

Future Functionality

• save resulting datasets and regression analyses
• manually delete all saved files at end of session
• algorithm tuning: hyperperameters adjusted (via dropdowns/checkboxes) to optimzie models
• predictive analysis: upload unlabeled data and predict dependent variable
• missing data (NaN) handling
• multiple header handling

Structure

• app.py is the server.
• index.html is the initial html page (includes upload function).
• ml.html is the machine learning html page (main interface).
• regressions.py runs regressions according to selected models and returns RMSE and R-Squared stats.
• ols_summary.py runs initial linear regression and initial random forest to generate coefficients, p-values and feature importances (to assist with feature/variable selection).
• global_y.py holds the y variable (for app.py).

Technologies Employed

• python
• numpy
• pandas
• flask
• jinja
• javascript
• jquery
• ajax
• bootstrap
• html
• css

resume-match