 
-- K-nearest-neighbour classifier in SQL --


-- create table: data samples --
DROP TABLE IF EXISTS DATAXY;
CREATE TABLE DATAXY (
  id INTEGER NOT NULL,    -- used as reference key --
  x REAL NOT NULL,        -- X data element --
  y REAL NOT NULL,        -- Y data element --
  cid INTEGER NOT NULL    -- class ID (known) --
);

-- create table: testing samples --
DROP TABLE IF EXISTS TESTXY;
CREATE TABLE TESTXY (
  id INTEGER NOT NULL,    -- used as reference key --
  x REAL NOT NULL,        -- X data element --
  y REAL NOT NULL,        -- Y data element --
  cid INTEGER            -- class ID (unknown) --
);


-- insert data samples --
INSERT INTO DATAXY VALUES ( 1, 1.0, -1.0, -1);  -- classID:'-1' --
INSERT INTO DATAXY VALUES ( 2, 2.0, -1.0, -1);
INSERT INTO DATAXY VALUES ( 3, 2.5,  1.0, -1);
INSERT INTO DATAXY VALUES ( 4, 3.0,  0.0, -1);
INSERT INTO DATAXY VALUES ( 5, 3.5,  2.0, -1);

INSERT INTO DATAXY VALUES ( 6, 11.0, -1.0, 1);  -- classID:'+1' --
INSERT INTO DATAXY VALUES ( 7, 12.0, -1.0, 1);
INSERT INTO DATAXY VALUES ( 8, 12.5,  1.0, 1);
INSERT INTO DATAXY VALUES ( 9, 13.0,  0.0, 1);
INSERT INTO DATAXY VALUES (10, 13.5,  2.0, 1);

-- insert testing samples --
-- Note: keep only one row active for N=1 testing --
INSERT INTO TESTXY VALUES (91, 13.5,  0.3, NULL);
INSERT INTO TESTXY VALUES (92, 6.5, -2.0, NULL);
INSERT INTO TESTXY VALUES (93, 16.5, -9.0, NULL);


-- k-nn fetch, simple version (works for N=1 test sample) --
-- Note: As it is, for N>1 the results of all testing
-- samples are merged into one table without grouping
-- i.e., some FOR-EACH structure is necessary for this.

SELECT 
  --id1, id2, cidk  -- display the top K neighbours
  id2, SUM(cidk)/ABS(SUM(cidk))  -- display only the result
FROM TESTXY, (
  SELECT 
    DATAXY.id AS id1, TESTXY.id AS id2,
    -- calculate the Euclidean distance (squared) --
    ((DATAXY.x-TESTXY.x) * (DATAXY.x-TESTXY.x)
    + (DATAXY.y-TESTXY.y) * (DATAXY.y-TESTXY.y)) AS dist,
    -- keep the coresponding class ID --
    DATAXY.cid AS cidk
  FROM DATAXY, TESTXY
  ORDER BY id2, dist ASC -- sort against distances
  LIMIT 5 -- this is the K parameter
);


-- display table sizes as separator from results below
SELECT COUNT(*) FROM DATAXY;
SELECT COUNT(*) FROM TESTXY;


-- k-nn fetch, simple version (works for N>1 test samples) --
-- Note: A few hacks are necessary in order to get precisely
-- the 'summary' result with only sign (not sums) as classID.

-- activate for full results:
--SELECT id1, id2, cidk, rnum from (
-- activate if SQL supports SIGN(.):
--SELECT id2, SIGN(ssum) FROM (
-- take distinct rows, avoid duplicates dur to (X,Y) pairing
SELECT DISTINCT 
  -- take the sum over each classID in the results
  id2, SUM(cidk) OVER(partition BY id2) AS ssum 
FROM (
  -- take distinct rows, avoid duplicates dur to (X,Y) pairing
  SELECT DISTINCT
    id1, id2, cidk, 
    -- add run numbering, resets for every classID (grouping)
    row_number() OVER(PARTITION BY id2) as rnum
  FROM TESTXY, (
    SELECT 
      DATAXY.id AS id1, TESTXY.id AS id2,
      -- calculate the Euclidean distance (squared) --
      ((DATAXY.x-TESTXY.x) * (DATAXY.x-TESTXY.x)
      + (DATAXY.y-TESTXY.y) * (DATAXY.y-TESTXY.y)) AS dist,
      -- keep the coresponding class ID --
      DATAXY.cid AS cidk
    FROM DATAXY, TESTXY
    ORDER BY id2, dist ASC 
  ) -- sort against distances
) 
WHERE rnum <= 5; -- this is the K parameter
--);  -- used when additional SELECT wrapper is enabled

created 2 years ago by ApneaCoding

SQLite online editor

Write, Run & Share SQLite queries online using OneCompiler's SQLite online editor and compiler for free. It's one of the robust, feature-rich online editor and compiler for SQLite. Getting started with the OneCompiler's SQLite editor is really simple and pretty fast. The editor shows sample boilerplate code when you choose language as 'SQLite' and start writing queries to learn and test online without worrying about tedious process of installation.

About SQLite

SQLite is an in-process C library that implements small, fast, serverless, zero-configuration, transactional SQL database engine.

Key Features:

Very small and light weight
Serverless
Free to use
Self contained as no other dependencies required.
zero config

Syntax help

Useful Commands

1. CREATE

CREATE TABLE table_name (
                column1 datatype,
                column2 datatype,
                ....);

Example

CREATE TABLE EMPLOYEE (
  empId INTEGER PRIMARY KEY,
  name TEXT NOT NULL,
  dept TEXT NOT NULL
);

2. ALTER

ALTER TABLE Table_name ADD column_name datatype;

Example

INSERT INTO EMPLOYEE VALUES (0001, 'Dave', 'Sales');

3. DROP

DROP TABLE table_name;

4. INSERT

INSERT INTO table_name (column1, column2, column3, ...) VALUES (value1, value2, value3, ...);

Note: Column names are optional.

Example

INSERT INTO EMPLOYEE VALUES (0001, 'Ava', 'Sales');

5. SELECT

SELECT column1, column2, ...
FROM table_name
[where condition];

Example

SELECT * FROM EMPLOYEE where dept ='sales';

6. UPDATE

UPDATE table_name
SET column1 = value1, column2 = value2, ...
WHERE condition;

Example

UPDATE EMPLOYEE SET dept = 'Sales' WHERE empId='0001';

7. DELETE

DELETE FROM table_name where condition;

Example

DELETE from EMPLOYEE where empId='0001';

8. CREATE INDEX

  CREATE [UNIQUE] INDEX index_name on table_name(column_name);

9. DROP INDEX

DROP INDEX index_name ON table_name;

10. Create a View

CREATE VIEW View_name AS 
Query;

11. How to call view

SELECT * FROM View_name;

12. Altering a View

ALTER View View_name AS 
Query;

13. Deleting a View

DROP VIEW View_name;

14. INNER JOIN

SELECT * FROM TABLE1 INNER JOIN TABLE2 where condition;

15. LEFT JOIN

SELECT * FROM TABLE1 LEFT JOIN TABLE2 ON condition;

16. RIGHT JOIN

SELECT * FROM TABLE1 RIGHT JOIN TABLE2 ON condition;

17. CROSS JOIN

SELECT select_list from TABLE1 CROSS JOIN TABLE2;