Roadmap for Geochemistry π

v0.6.0 -> v0.7.0 🎯

• Data Modeling

• CLI Pipeline

  • New Mode
    • Network analysis🌟
    • Statistical weight bootstrap resampling🌟
  • New algorithms
    • Local outlier factor algorithm in anomaly detection mode🌟
  • New application functions
    • Precision-recall vs. threshold diagram in classification mode🌟
  • Refactoring
    • Allow users to submit batch processing file
    • Use enum class to replace the comparability of string value
  • Others
    • Silence of dependency downloading when first launching
    • Run No. in the output data🌟
    • Download by one-click with input data in the desktop🌟

• Web Development

  • UI
    • Login/Register page
    • Product page
  • Frontend
    • Setup frontend code in Zhejiang University cloud
    • Login/Register page
  • Backend
    • Research and demo the CLI product in web portal
    • Improve login and register
  • Rate limiting
    • Change synchronization into asynchronization
      • Process pool
      • Message queue

• Testing

  • Exhaust all options automatically
  • Unit testing

• Documentation

  • Rewrite Add New Model to Framework

v0.5.0 -> v0.6.0 🏆

• Web Development

  • Single sign-on via DDE
  • UI design
    • Draft of training pipeline
    • Figma version
  • Web Page
    • Main Page
    • Introduction Page
    • Product Page
  • Storage Research - SQLAlchemyStore and FilesystemStore
    • Object Storage Service via Alibaba Cloud
    • MySQL via Alibaba Cloud
  • MLflow Integration
    • MLflow YML file error
    • Management of multiple users' resource using OSS and MySQL
  • CI/CD
    • Code formatting of js code is controlled by frontend

• Pipeline

  • The storage of the predicted value of the training data
  • Optimization of the loading of built-in application data
  • Dropping rows with missing values in selected column
  • Summary directory for all result
  • Prediction for the training set
  • Algorithm
    • Abnormal detection
      • Isolation forest algorithm addition
    • Decomposition
      • 2d scatter_diagram
      • Heatmap diagram
      • Contour diagram
    • Regression
      • Formula showing error fix in linear regression
      • Ridge regression addition
    • Clustering
      • Affinity propagation addition

• Documentation

  • Make developer-related documents
    • Append the corresponding operations on Windows in the developer section
  • Mind map of all options
  • Online docs layout mismatch
  • add citation info
  • update installation manual
  • update clustering algorithm example
  • add abnormal detection example

v0.4.0 -> v0.5.0 🏆

• Pipeline

  • Missing value process section
    • provide three ways to deal with missing values
      • Keep the missing values. Subsequently, only the models that support missing values are available.
      • Drop the rows with missing values
      • Impute the missing values with one of the imputation techniques.
  • Fixed random state for all models

• Algorithm & Function

  • Clustering & Decomposition
    • more common application functions
  • New algorithm
    • Bayesian ridge regression
    • Agglomerative clustering

• Documentation

  • Make developer-related documents
    • Algorithm addition procedure
    • functionality addition procedure

• Video

  • Machine Learning Lifecycle management

v0.3.0 -> v0.4.0 🏆

• Access Layer

  • Encapsulate a .exe software version

• Core Components

  • Model inference in customized ML pipeline (CLI version) by transform pipeline

• Documentation

  • Make design diagrams of the whole project

• Pipeline

  • Feature selection
  • CSV data file import
  • Data selection function with null, space and Chinese parentheses dection functionality
  • Feature scaling for unsupervised learning
  • Built-in inference dataset loading

• Algorithm

  • New algorithm
    • KNN classification
    • SGD classification
    • Gradient Boosting classification
    • SGD regression
    • Elastic Net Regression
  • Classification
    • Sample balance
    • Labels customization
    • Multi-class label and binary label training for all classification models
  • Decomposition
    • Reduced data storage for all decomposition model
  • Clustering
    • Silhouette score frequency diagram for all clustering model
    • Two clustering model score for all clustering model

• Enhancement

  • Lasso regression model with automatic parameter tuning functionality

v0.2.1 -> v0.3.0 🏆

• Functionality Building

  • Build T-SNE algorithm for decomposition section
  • Build advanced operation for feature engineering section
  • Build CLI command to activate offline server

• Access Layer

  • Build up web portal template
    • Login page
    • Home page
  • Build up RESTful APIs template
    • OAuth2 authorization
    • SQLite database

• ML Task Storage

  • Define and implement the schema to store metadata and artifact
    • Schema specification - MLflow
    • Implementation
      • Scikit-learn & MLflow
      • FLAML & MLflow
      • Ray & MLflow
    • Artifact
  • Formalize the local store format
    • Data Store
    • Image Store
    • Model Store

• Documentation

  • Migrate online document into organization repository
  • Optimize and integrate online documents
  • Optimize and resolve sphinx related problem
  • Make developer-related documents
    • Git operation
    • Local deployment
    • Docker deployment
  • Make framework-related design drawing
    • System architecture diagram
    • Customized ML pipeline diagram
    • Design pattern diagram
    • Workflow class diagram
    • Storage mechanism diagram
  • Algorithm & Function Progress Form

• Optimization

  • Abstract core functionalities to simplify the pipeline
    • Refactor four modes application functions
    • Build up Mixin class
  • Regulate CI for Python and Typescript
  • Regulate environment variable management
  • Fix statistic analysis bug
  • Improve World Map Projection by using dynamical importing