PROMPTS = {
"Analyze_database": """Please perform database analysis focusing on {database}.
Include in your analysis:
- List all tables in the database
- List all objects in the database
- Show detailed information about a database tables
Format your analysis in a well-structured manner with clear headings and concise explanations.
""",
"Analyze_table": """Provide all the details about {table} in database {database}.
Include in your analysis:
1. Show detailed information about a table {table} in database {database}
2. Check negative values in the table
3. Check missing values in the table
4. Check distinct values in the table for each column
4. Show mean and standard deviation for each numeric column in the table
Be comprehensive but focus on quality over quantity."
""",
"glm": """Provide all the details about table {table} in database {database}.
Include in your analysis:
1. Train, predict and evaluate GLM regression model
drop table train_test_out;
create multiset table train_test_out as (
select * from TD_TrainTestSplit (
on {table} as InputTable
USING
IDColumn ('id')
TrainSize (0.8)
TestSize (0.2)
Seed (17)
) as dt
) with data;
2. Create train and test tables
drop table weather_train;
create multiset table weather_train as (
select * from train_test_out where TD_IsTrainRow = 1
) with data;
create multiset table weather_test as (
select * from train_test_out where TD_IsTrainRow = 0
) with data;
3. Scale the train and test data
drop table fit_out_table;
put all numeric columns from weather_train in the TargetColumns() function
create multiset table fit_out_table as (
select * from TD_ScaleFit (
on weather_train as InputTable
USING
TargetColumns ()
ScaleMethod ('STD')
) as dt
) with data;
put all numeric columns from weather_train in the Accumulate() function
drop table weather_train_scaled;
create multiset table weather_train_scaled as (
select * from TD_ScaleTransform (
on weather_train as InputTable
on fit_out_table as FitTable DIMENSION
USING
Accumulate()
) as dt)
with data;
put all numeric columns from weather_test in the Accumulate() function
drop table weather_test_scaled;
create multiset table weather_test_scaled as (
select * from TD_ScaleTransform (
on weather_test as InputTable
on fit_out_table as FitTable DIMENSION
USING
Accumulate('[1]','[3]','[9]','[11:12]','[23:25]')
) as dt)
with data;
drop table fit_out_table;
create multiset table fit_out_table as (
select * from TD_OneHotEncodingFit (
on weather_train_scaled as InputTable
USING
TargetColumn ('WindGustDir','WindDir9am','WindDir3pm','RainToday','RainTomorrow')
IsInputDense('true')
Approach('auto')
CategoryCounts(16,16,16,2,2)
) as dt
) with data;
put all numeric columns from weather_train in the TargetColumns() function
drop table weather_train_s;
create multiset table weather_train_s as (
select * from TD_OneHotEncodingTransform (
on weather_train_scaled as InputTable
on fit_out_table as FitTable DIMENSION
USING
IsInputDense('true')
) as dt)
with data;
create multiset table weather_test_s as (
select * from TD_OneHotEncodingTransform (
on weather_test_scaled as InputTable
on fit_out_table as FitTable DIMENSION
USING
IsInputDense('true')
) as dt)
with data;
4. Train a GLM model on train data. put all numbering columns from weather_train_s in the InputColumns() function
drop table glm_model;
create multiset table glm_model as (
select * from TD_GLM (
ON weather_train_s
USING
InputColumns('')
ResponseColumn('rainfalltomorrow')
Family('Gaussian')
) as dt
) with data;
5. Predict test data using GLM model.
drop table glm_predict_out;
create multiset table glm_predict_out as (
select * from TD_GLMPredict (
ON weather_test_s as InputTable
ON glm_model as ModelTable DIMENSION
USING
IDColumn('id')
Accumulate('rainfalltomorrow')
) as dt
) with data;
6. Evaluate GLM model by computing MSE and R2 values
select * from TD_RegressionEvaluator (
on glm_predict_out_modified as InputTable
USING
ObservationColumn('rainfalltomorrow')
PredictionColumn('prediction')
Metrics('MSE','R2')
) as dt;
"""
}
