Final Exam: Advanced Analytics and Machine Learning in Snowflake

WHAT YOU WILL LEARN

• 

  Identify snowflake ai/ml offerings across both snowflake cortex and snowflake ml and the functionality available across the platform
  

  outline support in snowpark ml for model training using scikit-learn, xgboost, and lightgbm, as well as for hyperparameter tuning
  

  connect to snowflake from jupyter and use the snowpark api from python
  

  utilize the snowflake ml apis to compute correlation matrices, construct pipelines, and fit models
  

  recognize the process and benefits of registering a model version
  

  register models and versions, view model artifacts, delete model versions, and invoke model methods dynamically
  

  provide an overview of the intuition behind clustering depth and the number of overlapping partitions
  

  create a snowflake ml pipeline for logistic regression
  

  create tags and associate metrics and a tag with a model
  

  utilize snowpark-optimized warehouses and grid search for hyperparameter tuning
  

  register a model with tuned hyperparameter values with the model registry
  

  identify the need for the snowflake feature store and analyze how feature views and entities work
  

  create a feature store and entity using snowpark apis
  

  create a managed feature view and analyze the implementation as a dynamic table
  

  create an external feature view, join two feature views, and query for all feature views associated with a specific entity
  

  create a basic streamlit app to display data from a table when a checkbox is selected
  

  add heatmaps, scatter plots, and other seaborn and matplotlib visualizations to a streamlit app
  

  implement sliders, selection boxes, radio buttons, and other ui controls in streamlit
  

  access the model registry and display a dropdown of models and model version numbers in a streamlit app
  

  share the completed streamlit app with a different user in view-only mode
  

  outline the steps for training an anomaly detection model and how to use it for prediction
  

  recognize how to analyze each column in the output of the anomaly detection model
• 

  use snowflake ml functions to train a single-series unsupervised anomaly detection model
  

  invoke the anomaly detection function, interpret the results, and save them to a table
  

  create an anomaly detection model for single-series data with no exogenous variables using snowflake ml functions and a filtered query on a multi-series dataset
  

  extend the single-series anomaly detection model by adding exogenous variables and observe changes in model sensitivity and feature importance scores
  

  save model results to a table using sqlid and the result_scan function, then calibrate model sensitivity using prediction_interval
  

  extend the anomaly detection model to work with multi-series data and verify that model feature scores are now reported for each series
  

  invoke the explain_feature_importance and show_evaluation_metrics on a time series forecasting model and analyze the results
  

  extend time series forecasting models by providing exogenous explanatory variables
  

  utilize snowpark ml functions for multi time series forecasting
  

  generate sql code that builds and invokes a time series forecasting model using the ai & ml studio wizard for forecasting
  

  analyze and execute the sql code generated by the ai & ml studio from the forecasting workflow
  

  generate sql code that invokes ml classification functions using the ai & ml studio wizard for classification
  

  analyze and execute the sql code generated by the ai & ml studio from the classification workflow
  

  analyze evaluation metrics, global evaluation metrics, and feature importance scores from the output of models created by snowflake ai & ml studio
  

  recognize how to use temperature and top_p hyperparameters to determine the predictability of llm output
  

  use the complete cortex llm function with different types of roles
  

  utilize the temperature, max_p, and guardrails properties to control the attributes of responses from the complete cortex llm function
  

  invoke the extract_answer, summarize, sentiment, and translate functions from sql
  

  invoke the complete, extract_answer, summarize, sentiment, and translate functions from python
  

  recognize the functionality of snowflake copilot, universal search, and document ai
  

  implement search optimizations for individual columns for both equality and substring matches
  

  outline retrieval augmented generation (rag) and the use of cortex search for rag