Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Tour of Apache PredictionIO in 10 Minutes

171 views

Published on

Scala eXchange 2017

  • Be the first to comment

  • Be the first to like this

Tour of Apache PredictionIO in 10 Minutes

  1. 1. Tour of Apache PredictionIO in 10 Minutes Scala eXchange Dec 14th, 2017
  2. 2. Web Developer We'd like to give users recommendations.
  3. 3. App Give us items, please. Web Developer Here you are.
  4. 4. Give us items, please. Web Developer Here you are. App It's easy for you to say.
  5. 5. Prediction query result App
  6. 6. Training Data Prediction query result collect some events App
  7. 7. Training Data Predictive Model Prediction train query result App collect some events
  8. 8. Training Data Predictive Model Prediction train deploy query result App collect some events
  9. 9. Training Data query result App collect some events PredictionIO
  10. 10. Overview ● Open source machine learning framework ○ Building an end-to-end machine learning platform ● Built-in storage facilities ○ Training data: (e.g.: Elasticsearch, HBase) ○ Models: (e.g.: HDFS, S3)
  11. 11. Apply Governance ● Involve personal efforts ○ Inconsistent architectures ● Apply governance using PIO ○ Improve efficiency and productivity
  12. 12. Training Data Predictive Model Prediction train deploy query result App collect some events
  13. 13. Training Data Predictive Model Prediction deploy query result App collect some events predict train
  14. 14. Templates ● All kinds of machine learning tasks ○ Template gallery offers templates in advance ○ Use by downloading a template ● Customizable
  15. 15. Process
  16. 16. Process git clone choose a template collect some events Ap
  17. 17. Process git clone collect some events pio build choose a template
  18. 18. Process git clone collect some events pio build pio train store model choose a template
  19. 19. Process git clone collect some events pio build pio train pio deploy store model choose a template
  20. 20. Process git clone collect some events pio build pio train pio deploy App query store model choose a template
  21. 21. Process git clone collect some events pio build pio train pio deploy App query result store model choose a template Note that some steps are left out for brevity.
  22. 22. Algorithm def train(sc: SparkContext, data: PreparedData): Model = { // Returns a trained model } def predict(model: Model, query: Query): PredictedResult = { // Returns a prediction }
  23. 23. Algorithm def train(sc: SparkContext, data: PreparedData): Model = { // Returns a trained model } def predict(model: Model, query: Query): PredictedResult = { // Returns a prediction } Produce a model from prepared data
  24. 24. Algorithm def train(sc: SparkContext, data: PreparedData): Model = { // Returns a trained model } def predict(model: Model, query: Query): PredictedResult = { // Returns a prediction } Produce a prediction from a query and trained model
  25. 25. Community ● Support : user@predictionio.apache.org ● https://predictionio.apache.org

×