scientist

A practical toolbox for machine learning and statistical analysis

Install

• currently only support mac os
• download the whole folder, and open terminal
• install pipenv and run pipenv install and pipenv shell
• then run python app.py

Features

load data

• now the app takes two types of data:
  • table data: csv file with header
  • text data: txt file

views

Table View

• table view is used to look at the origin data
• large data frame is split with pagination
• to get a more general view, you can shuffle the data
• basic statics and correlation are also available

Plot View

• plot view will give a more abstract idea of the dataset through plots
• you can select which features you want to plot
• currently support: kde, histogram, line, box, bar and scatter matrix

Text View

• this is only for text data, we can read through the content

preprocess

• we provide several scalers to scale the features
• popular encoders
• when the dataset is too large, drop columns, rows or just draw a sample
• different strategies to handle missing data

clustering

• build clusters using the following algorithms
  • kmeans
  • hierarchical methods
  • affinity propagation
• visualize the clusters using popular plot methods
  • t-sne
  • radviz
  • top 2 pcas
  • hierarchical tree(only for hierarchical methods)

classification

• support both biclass problems and multiclass problems
• lots of common classifiers
• train test split and cross validation
• basic metrics and confusion matrix

regression

• linear regression with basic statistics
• glm (gamma)

coincidence