Using regr_slope in sql query

Question

Hi ,
I have one question on predicting the sale for next 5 years  along with month  the data as follows.
YEAR  Month  Total_Sale
2009    1          88187
2009     2         87654
2009     3         87656
2008     1         10000
2008     2          30000
2008     3          40000
How to use regr_slope and other function for the prediction of sale to next year?
Thanks,
HS

Answer 1

Hi,
de45e462-4cf1-4ad2-8f0b-4688375a6400 wrote:
Hi ,
Fixed the input data script as follows as per the suggestion.
REM INSERTING into SALE_GROWTH
SET DEFINE OFF;
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2009,1,1000);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2009,2,2000);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2009,3,3000);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2009,4,1000);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2009,5,5000);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2009,6,2000);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2009,8,3000);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2008,9,700);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2008,10,10000);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2008,11,30000);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2008,12,2000);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2008,9,6000);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2008,8,100);
Now the output  i want for the future years   2014,2015.
-HS
What is the output you want for the future years 2104, 2015?
It seems like this sample data isn't really any simpler than the old sample data.  Again, why is month 4 (I assume that's April, but it doesn't really matter) lower than all the other months, before and after it?
It's unclear if you're really talking about linear regression, or something more complicated.  Do you really need to post sample data that's more complicated than this?
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2009,1,1000);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2009,2,2000);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2009,3,3000);
Insert into SALE_GROWTH (YEAR,MONTH,TOTAL_SALE) values (2009,5,5000);
Here, there is a predictable increase of exactly 1000 every single month, so we know the expected results  for the next 12 months would be:
MONTH_YEAR      EXPECTED_SALES
September, 2013          57000
October, 2013            58000
November, 2013           59000
December, 2013           60000
January, 2014            61000
February, 2014           62000
March, 2014              63000
April, 2014              64000
May, 2014                65000
June, 2014               66000
July, 2014               67000
August, 2014             68000
Here's one way to get those results, using the aggregate REGR_SLOPE and REGR_INTERCEPT functions:
WITH   got_slope_and_intercept   AS
    SELECT   REGR_SLOPE     (total_sale, month + (12 * year))   AS slope
    ,        REGR_INTERCEPT (total_sale, month + (12 * year))   AS intercept
     FROM   sale_growth
SELECT   TO_CHAR ( ADD_MONTHS (SYSDATE, LEVEL)
                 , 'fmMonth, YYYY'
                 )       AS month_year
,        intercept + ( slope * (        EXTRACT (MONTH FROM SYSDATE)
                               + ( 12 * EXTRACT (YEAR  FROM SYSDATE))
                               + LEVEL
                     )   AS expected_sales
FROM     got_slope_and_intercept
CONNECT BY   LEVEL <= 12