Machine Learning in Autonomous Data Warehouse
0 likes419 views
The document outlines the features and direction of Oracle's Machine Learning capabilities within its Autonomous Database, highlighting various machine learning techniques such as regression, classification, clustering, and anomaly detection. It emphasizes how machine learning can automate traditional database maintenance tasks, enabling database administrators to focus on innovation. Additionally, it discusses the evolution of data roles and the importance of leveraging data for predictive applications and analytics.
Machine Learning in Autonomous Data Warehouse
- 1. VP AIOps for the Autonomous Database Sandesh Rao Machine Learning in Autonomous Data Warehouse @sandeshr https://www.linkedin.com/in/raosandesh/ https://www.slideshare.net/SandeshRao4
- 2. The following is intended to outline our general product direction. It is intended for information purposes only, and may not be incorporated into any contract. It is not a commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions. The development, release, timing, and pricing of any features or functionality described for Oracle’s products may change and remains at the sole discretion of Oracle Corporation. Statements in this presentation relating to Oracle’s future plans, expectations, beliefs, intentions and prospects are “forward-looking statements” and are subject to material risks and uncertainties. A detailed discussion of these factors and other risks that affect our business is contained in Oracle’s Securities and Exchange Commission (SEC) filings, including our most recent reports on Form 10-K and Form 10-Q under the heading “Risk Factors.” These filings are available on the SEC’s website or on Oracle’s website at http://www.oracle.com/investor. All information in this presentation is current as of September 2019 and Oracle undertakes no duty to update any statement in light of new information or future events. Safe harbor statement
- 3. 1. Overview of ML and the Autonomous Database 2. Regression 3. Classification 4. Clustering 5. Anomaly detection 6. Workload prediction 7. Dynamic maintenance windows 8. Oracle Machine Learning examples Agenda
- 4. Overview of ML and Autonomous Database
- 5. Tasks Specific to Business and Innovation • Architecture, planning, data modeling • Data security and lifecycle management • Application related tuning • End-to-End service level management Maintenance Tasks • Configuration and tuning of systems, network, storage • Database provisioning, patching • Database backups, H/A, disaster recovery • Database optimization Traditionally DBAs are Responsible for: Value Scale Innovation Maintenance
- 6. Tasks Specific to Business and Innovation • Architecture, planning, data modeling • Data security and lifecycle management • Application related tuning • End-to-End service level management Maintenance Tasks • Configuration and tuning of systems, network, storage • Database provisioning, patching • Database backups, H/A, disaster recovery • Database optimization Freedom from Drudgery for DBA: More Time to Innovate and Improve the Business Autonomous Database Removes Generic Tasks Value Scale Innovation Maintenance
- 7. Machine Learning Solving data-driven problems Discovering insights Making predictions Data Security Data classification, Data life-cycle mgmt Application Tuning SQL tuning, connection mgmt The Evolution of the DBA/Database Developer Role Data Engineer Architecture, “data wrangler”
- 8. Data extraction Data wrangling Deriving new attributes (“feature engineering”) … … … Import predictions & insights Translate and deploy ML models Automate You Are Probably Already Doing Most of This Work! Database Developer to Data Scientist Journey 1 - https://www.infoworld.com/article/3228245/data-science/the-80-20-data-science-dilemma.html Typically 80% of the work Most data scientists spend only 20 percent of their time on actual data analysis and 80 percent of their time finding, cleaning, and reorganizing huge amounts of data, which is an inefficient data strategy1 Eliminated or minimized with Oracle Data Management platform becomes combine/hybrid DM + machine learning platform
- 9. Albert Einstein “If I had an hour to solve a problem I'd spend 55 minutes thinking about the problem and 5 minutes thinking about solutions.”
- 10. Lots of Data needs to be crunched • No time to manually sift through the data Machine Learning has become accessible • Anyone can be a Data Scientist • Software and algorithms are available • Frameworks allow for massive training with no coding • CI/CD available for MLOps Business use cases - How to find value from the data Why Machine Learning for us and why now?
- 11. Analytics Value vs. Maturity Reports & Dashboards Data Information Predictions & Insights Appls with ML Analytical Maturity ValueofAnalytics Diagnostic Analysis & Reports Predictive / Machine Learning “ML Enabled” Applications What Happened? Why it Happened? What WILL happen? Automated ML Appls
- 12. Database Developer to Data Scientist Journey
- 13. ML Project Workflow Set the business objectives Gather compare and clean data Identify and extract features (important columns) from imported data This helps us identify the efficiency of the algorithm Take the input data which is also called the training data and apply the algorithm to it For the algorithm to function efficiently, it is important to pick the right value for hyper parameters (algorithm input parameters to the algorithm) Once the training data in the algorithm are combined we get a model 1 2 3 4 5
- 14. Types of Machine Learning Supervised Learning Predict future outcomes with the help of training data provided by human experts Semi-Supervised Learning Discover patterns within raw data and make predictions, which are then reviewed by human experts, who provide feedback which is used to improve the model accuracy Unsupervised Learning Find patterns without any external input other than the raw data Reinforcement Learning Take decisions based on past rewards for this type of action
- 15. Regression
- 16. REGRESSION Predicting numbers Customer lifetime value Estimate optimal pricing House price estimates - 10 0 10 20 30 40 50 60
- 17. Classification
- 18. CLASSIFICATION Membership of a known class Identify likely high value customers Find customer likely to churn Fraud detection
- 19. Clustering
- 20. CLUSTERING Membership of an inferred class Customer segmentation Credit risk evaluation Document similarity
- 22. ANOMALY DETECTION Outliers Dentist billing 85 fillings / hour Employees with high claims / grade One variable moving out of sync
- 23. ASSOCIATION RULES Finding like- minded people You might be interested in... Root cause analysis Identify “harbingers of failure”
- 24. TIME SERIES Temporal Aspect Hitting a threshold Forecasting energy use Seasonality of data
- 25. GRAPH ANALYTICS Supplement ML algorithm Customer churn Network outage Fraud detection
- 26. NEURAL NETWORKS Learn (More) Like A Human Classification Regression Deep learning http://neuralnetworksanddeeplearning.com/chap1.html
- 27. Machine Learning Algorithms • Multiple Regression, Support Vector Machine, Linear Model, LASSO, Random Forest, Ridge Regression, Generalized Linear Model, Stepwise Linear Regression Regression Association & Collaborative Filtering Reinforcement Learning - brute force, Monte Carlo, temporal difference.... • Many different use cases Neural network & deep Learning with Deep Neural Network • Hierarchical k-means, Orthogonal Partitioning Clustering, Expectation- Maximization Clustering Feature Extraction/Attribute Importance / Component Analysis • Decision Tree, Naive Bayes, Random Forest, Logistic Regression, Support Vector Machine Classification
- 28. ML To determine workload and deviation from it
- 29. What is Workload Automatically check workload for past x mins Decide if workload is abnormally high Highlight any abnormal workload issues Optionally run on demand Optionally snooze checking of a component Calculated via machine learning
- 30. Adaptive Learning Workload Process Captures metrics for key performance dimensions across 5 X 1 minute time windows CAPTURE1 Using semi-supervised learning via SME threshold rules, the following models are retrained : • Isolation Forest • One-Class Support Vector Machine • Local Outlier Factor Model with highest confidence becomes the primary, if confidence is high enough TRAIN2 Straight after capture, the primary model is used to predict anomalies. Where anomalies are identified, metrics are compared to SME threshold rules to identify the type of anomaly PREDICT3 Every 5 Mins Every Week Every 5 Mins
- 31. Prediction (Every 5 minutes) 5 X 1 min metrics captured for each dimension & ASH report captured for later analysis Metrics evaluated by the primary model to determine if there are anomalies If there is no primary model (i.e. <7 days of data or <=95% model confidence) then SME rules are used for anomaly detection Each anomaly is compared against the SME rules to determine which dimension it applies to Any anomalies are raised along with recently captured ASH report
- 32. Resource usage prediction Configurable threshold boundary – notify Admin of forecasts above here Actual values (Black) Forecast values (Blue line) Upper & lower forecast range (light blue area) Unusual values (anomalies) Future forecast values
- 33. ML To determine dynamic maintenance windows
- 34. Identify Relevant Workload Metrics • Ex: Average Active Sessions, CPU/Mem/IO Utilization Time Series Decomposition • Trend • Seasonality • Residual Workload Seasonality Determination Locating Minimas Optimum Window Identification and Validation Model Generation and Training Flow Maintenance Slot Identification
- 35. Maintenance window identification START_TIME CNT 2018-04-11 15:00:00 290 2018-04-11 16:00:00 31120 2018-04-11 17:00:00 21530 2018-04-11 18:00:00 26240 2018-04-11 19:00:00 40520 2018-04-11 20:00:00 54270 2018-04-11 21:00:00 51460 2018-04-11 22:00:00 44310 2018-04-11 23:00:00 25690 START_TIME 2018-04-11 15:00:00 -0.226098 2018-04-11 16:00:00 -0.069821 2018-04-11 17:00:00 -0.350088 2018-04-11 18:00:00 -0.187483 2018-04-11 19:00:00 -0.513240 2018-04-11 20:00:00 0.019737 2018-04-11 21:00:00 0.059213 2018-04-11 22:00:00 -0.011312 2018-04-11 23:00:00 -0.179156 START_TIME 2018-04-11 15:00:00 5.669881 2018-04-11 16:00:00 10.345606 2018-04-11 17:00:00 9.977203 2018-04-11 18:00:00 10.175040 2018-04-11 19:00:00 10.609551 2018-04-11 20:00:00 10.901727 2018-04-11 21:00:00 10.848560 2018-04-11 22:00:00 10.698966 2018-04-11 23:00:00 10.153857 Current Date : 2018-05-12 15:00:00 Current Position in Seasonality : -0.22609829742533585 Best Maintenance Period in next Cycle : 2018-05-12 19:00:00 Worst Maintenance Period in next Cycle : 2018-05-13 08:00:00 Original observation data 1 Convolution filter & average 2 Calculate seasonality 3 Use seasonality to predict best maintenance window 4
- 36. OML examples
- 37. Simple SQL Syntax—Statistical Comparisons (t-tests) Compare AVE Purchase Amounts Men vs. Women Grouped_By INCOME_LEVEL Statistical Functions SELECT SUBSTR(cust_income_level, 1, 22) income_level, AVG(DECODE(cust_gender, 'M', amount_sold, null)) sold_to_men, AVG(DECODE(cust_gender, 'F', amount_sold, null)) sold_to_women, STATS_T_TEST_INDEPU(cust_gender, amount_sold, 'STATISTIC', 'F') t_observed, STATS_T_TEST_INDEPU(cust_gender, amount_sold) two_sided_p_value FROM customers c, sales s WHERE c.cust_id = s.cust_id GROUP BY ROLLUP(cust_income_level) ORDER BY income_level, sold_to_men, sold_to_women, t_observed; STATS_T_TEST_INDEPU (SQL) Example; P_Values < 05 show statistically significantly differences in the amounts purchased by men vs. women
- 38. Simple SQL Syntax—Attribute Importance - ML Model Build (PL/SQL) OAA Model Build and Real-time SQL Apply Prediction BEGIN DBMS_DATA_MINING.CREATE_MODEL( model_name => 'BUY_INSURANCE_AI', mining_function => DBMS_DATA_MINING.ATTRIBUTE_IMPORTANCE, data_table_name => 'CUST_INSUR_LTV', case_id_column_name => 'cust_id', target_column_name => 'BUY_INSURANCE', settings_table_name => 'Att_Import_Mode_Settings'); END; / SELECT attribute_name, rank , attribute_value FROM BUY_INSURANCE_AI ORDER BY rank, attribute_name; Model Results (SQL query) ATTRIBUTE_NAME RANK ATTRIBUTE_VALUE BANK_FUNDS 1 0.2161 MONEY_MONTLY_OVERDRAWN 2 0.1489 N_TRANS_ATM 3 0.1463 N_TRANS_TELLER 4 0.1156 T_AMOUNT_AUTOM_PAYMENTS 5 0.1095 A1A2A3A4 A5A6 A7
- 39. Key Features Collaborative UI for data scientists • Packaged with Autonomous Data Warehouse Cloud • Easy access to shared notebooks, templates, permissions, scheduler, etc. • SQL ML algorithms API • Supports deployment of ML analytics Machine Learning Notebook for Autonomous Data Warehouse Cloud Oracle Machine Learning
- 40. Oracle Machine Learning Key Features: • Collaborative UI for data scientist and analysts • Packaged with Autonomous Databases • Quick start Example notebooks • Easy access to shared notebooks, templates, permissions, scheduler, etc. • OML4SQL • OML4Py coming soon • Supports deployment of OML models Machine Learning Notebooks included in Autonomous Databases Copyright © 2020 Oracle and/or its affiliates.
- 41. Oracle Machine Learning Key Features: • Collaborative UI for data scientist and analysts • Packaged with Autonomous Databases • Quick start Example notebooks • Easy access to shared notebooks, templates, permissions, scheduler, etc. • OML4SQL • OML4Py coming soon • Supports deployment of OML models Machine Learning Notebooks included in Autonomous Databases Copyright © 2020 Oracle and/or its affiliates.
- 42. Oracle Machine Learning for R / Python Transparency layer ‐ Leverage proxy objects so data remain in database ‐ Overload native functions translating functionality to SQL ‐ Use familiar R/Python syntax to manipulate database data Parallel, distributed algorithms ‐ Scalability and performance ‐ Exposes in-database algorithms available from OML4SQL Embedded execution ‐ Manage and invoke R or Python scripts in Oracle Database ‐ Data-parallel, task-parallel, and non-parallel execution ‐ Use open source packages to augment functionality OML4Py, Automated Machine Learning - AutoML ‐ Feature selection, model selection, hyper-parameter tuning Multiple Components/APIs of Oracle Machine Learning Database Server Client SQL Interfaces SQL*Plus SQLDeveloper OML4Py OML4R Copyright © 2020 Oracle and/or its affiliates. * Coming soon
- 43. Multiple Languages UIs Supported for End Users & Apps Development Oracle Machine Leaning Application DevelopersDBAs R & Python Data Scientists “Citizen” Data ScientistsNotebook Users & DS Teams New! New!
- 44. Coming Soon! | AutoML – new with OML4Py Auto Feature Selection – Reduce # of features by identifying most predictive – Improve performance and accuracy Increase data scientist productivity – reduce overall compute time Auto Algorithm Selection Much faster than exhaustive search Auto Feature Selection De-noise data and reduce # of features AutoTune Significant accuracy improvement Auto Algorithm Selection – Identify in-database algorithm that achieves highest model quality – Find best algorithm faster than with exhaustive search Auto Tune Hyperparameters – Significantly improve model accuracy – Avoid manual or exhaustive search techniques Copyright © 2020 Oracle and/or its affiliates. Enables non-expert users to leverage Machine Learning Data Table ML Model
- 45. Oracle Data Miner UI Easy to use to define analytical methodologies that can be shared SQL Developer Extension Workflow API and generates SQL code for immediate deployment Drag and Drop, Workflows, Easy to Use UI for “Citizen Data Scientist” Copyright © 2020 Oracle and/or its affiliates.
- 46. Target “best” customers who have GOOD CREDIT and make payments Business Usecase
- 47. Define Problem Statement Poorly Defined Better Data Mining Technique Predict employees that leave • Based on past employees that voluntarily left: • Create New Attribute EmplTurnover à O/1 Predict customers that churn • Based on past customers that have churned: • Create New Attribute Churn à YES/NO Target “best” customers • Recency, Frequency Monetary (RFM) Analysis • Specific Dollar Amount over Time Window: • Who has spent $500+ in most recent 18 months How can I make more $$? • What helps me sell soft drinks & coffee? Which customers are likely to buy? • How much is each customer likely to spend? Who are my “best customers”? • What descriptive “rules” describe “best customers”? How can I combat fraud? • Which transactions are the most anomalous? • Then roll-up to physician, claimant, employee…
- 48. Data loading and Review
- 49. Target “best” customers who have GOOD CREDIT and make payments Data Scoping
- 50. Create New Derived Attributes or “Engineered Features” Feature Engineering Source Attribute New Attribute/”Engineered Feature” Date of Birth AGE Address DISTANCE_TO_DESTINATION COMMUTE_TIME Call detail records (CDRs) #_DROPPED_CALLS PERCENT_iNTERNATIONAL Salary PERCENT_VS_PEERS Purchases TOTALS_PER_CATEGORY (e.g. Food, Clothing)
- 51. Create new derived attributes to tease more information out of the data. For example: • RECENCY, FREQUENCY, MONETARY (RFM Analysis) Create New Derived Attributes or “Engineered Features” Feature Engineering
- 52. Data remains in Database • Model building and scoring occur in- database • Leverage investment in Oracle IT • Eliminate data duplication - Eliminate separate analytical servers Deliver enterprise-wide “predictive” applications Don’t move the Data Traditional ML Hours, Days or Weeks Data Extraction Data Prep & Transformation Data Mining Model Building Data Mining Model “Scoring” Data Prep. & Transformation Data Import avings Model “Scoring” Embedded Data Prep Data Preparation Model Building Oracle’s in-DB Machine Learning Secs, Mins or Hours ORACLE AUTONOMOUS DATABASE
- 53. Increasing sources of relevant data can boost model accuracy More Data Variety—Better Predictive Models Model with 20 variables Model with “Big Data” and hundreds -- thousands of input variables including: • Demographic data • Purchase POS transactional data • “Unstructured data”, text & comments • Spatial location data • Long term vs. recent historical behavior • Web visits • Sensor data • etc. Naïve Guess or Random 100% 0% Population Size Responders Model with 75 variables Model with 250 variables 100% Engineered Features – Derived attributes/variable that reflect domain knowledge—key to best models
- 54. First, Identify the Key Attributes That Most Influence the Target Attribute Modeling and Machine Learning Attribute Importance Model
- 55. Next, Build Predictive Models to Predict Customers who are Likely to Have Good_Credit Modeling and Machine Learning Split Data into Train and Test Build and Test Classification Model
- 56. Test the ML model’s accuracy • Randomly selected “hold out” sample of data that was used to train the ML model • Compute Cumulative Gains, Lift, Accuracy, etc. • Review the attributes used in the model and model coefficients • Make sure the model makes sense Next, Build Predictive Models to Predict Customers who are Likely to Have Good_Credit Model Evaluation (Machine Learning) Model Evaluation
- 57. Simple SQL Apply scripts run 100% inside the Database for immediate ML model deployment Apply the Models to Predict “Best Customers” Deployment Model Apply/”Scoring”
- 59. Congratulations! Almost there J Data Scientist
- 60. Manage and Analyze All Your Data Big Data SQL / R SQL / R / Python Object Store “Engineered Features” – Derived attributes that reflect domain knowledge—key to best models e.g.: • Counts • Totals • Changes over time Boil down the Data Lake Architecturally, lots of options and flexibility
- 61. In-Database Machine Learning More Models Better Models Faster, More Secure Less Cost Ready to Deploy! No Need To Extract and Move Data Data stays in Database Zero time required. No production impact. Data Preparation and Transformation Accelerated with Automatic Data Prep No separate environment required. Much faster data prep. Data stays protected and secured. Data Mining and Model Building SQL, R, Python Oracle Data Miner UI OML Notebooks Oracle Data Miner and AutoML greatly speed model building. Less skill required. No coding. No Need to Transform Production Data Embedded Data Preparation No need for second production instance. Model Scoring Accelerated Via Exadata Database Machine Faster model validation Easy to repeat model building as often as needed
- 62. • OAA (Oracle Data Mining + Oracle R Enterprise) and ORAAH combined • OAA includes support for Partitioned Models, Transactional, Unstructured, Geo-spatial, Graph data. etc, Oracle’s Machine Learning & Adv. Analytics Algorithms CLASSIFICATION • Naïve Bayes • Logistic Regression (GLM) • Decision Tree • Random Forest • Neural Network • Support Vector Machine • Explicit Semantic Analysis CLUSTERING • Hierarchical K-Means • Hierarchical O-Cluster • Expectation Maximization (EM) ANOMALY DETECTION • One-Class SVM TIME SERIES • State of the art forecasting using Exponential Smoothing • Includes all popular models e.g. Holt-Winters with trends, seasons, irregularity, missing data REGRESSION • Linear Model • Generalized Linear Model • Support Vector Machine (SVM) • Stepwise Linear regression • Neural Network • LASSO * ATTRIBUTE IMPORTANCE • Minimum Description Length • Principal Comp Analysis (PCA) • Unsupervised Pair-wise KL Div • CUR decomposition for row & AI ASSOCIATION RULES • A priori/ market basket PREDICTIVE QUERIES • Predict, cluster, detect, features SQL ANALYTICS • SQL Windows, SQL Patterns, SQL Aggregates FEATURE EXTRACTION • Principal Comp Analysis (PCA) • Non-negative Matrix Factorization • Singular Value Decomposition (SVD) • Explicit Semantic Analysis (ESA) TEXT MINING SUPPORT • Algorithms support text • Tokenization and theme extraction • Explicit Semantic Analysis (ESA) for document similarity STATISTICAL FUNCTIONS • Basic statistics: min, max, median, stdev, t-test, F-test, Pearson’s, Chi-Sq, ANOVA, etc. R PACKAGES • CRAN R Algorithm Packages through Embedded R Execution • Spark MLlib algorithm integration EXPORTABLE ML MODELS • REST APIs for deployment X 1 X 2 A 1 A 2 A 3 A 4 A 5 A 6 A 7
- 63. ANALYTICAL SQL • SQL Windows • SQL Aggregate functions • LAG/LEAD functions • SQL for Pattern Matching • Additional approximate query processing: APPROX_COUNT , APPROX_SUM, APPROX_RANK • Regular Expressions • Linear regression • ANOVA (Analysis of variance) • Test Distribution fit (e.g. Normal distribution test, Binomial test, Weibull test, Uniform test, Exponential test, Poisson test, etc.) • Statistical Aggregates (min, max, mean, median, stdev, mode, quantiles, plus x sigma, minus x sigma, top n outliers, bottom n outliers) STATISTICAL FUNCTIONS • Descriptive statistics (e.g. median, stdev, mode, sum, etc.) • Hypothesis testing (t-test, F-test, Kolmogorov- Smirnov test, Mann Whitney test, Wilcoxon Signed Ranks test • Correlations analysis (parametric and nonparametric e.g. Pearson’s test for correlation, Spearman's rho coefficient, Kendall's tau-b correlation coefficient) • Ranking functions • Cross Tabulations with Chi-square statistics| Oracle’s Machine Learning & Adv. Analytics Algorithms
- 64. Algorithms Operate on Data ML and AI are just “Algorithms” Move the Algorithms; Not the Data!; It Changes Everything!
- 65. Thank You Any Questions ? Sandesh Rao VP AIOps for the Autonomous Database @sandeshr https://www.linkedin.com/in/raosandesh/ https://www.slideshare.net/SandeshRao4