We introduced a novel framework for explanation searching, SDEcho, which addresses the challenge of explaining aggregated sequence differences. SDEcho employs pruning techniques from three perspectives โ pattern, order, and dimension โ to ensure concise and accurate explanations while considering the characteristics of each pruning algorithm for hybrid optimization.
-
baselines/
- contains all pruning method baselines mentioned in the paper. -
benchmark/
- contains all the files necessary to the implementation related to benchmark generation and explanation confidence evaluation mentioned in the paper. -
data/
- contains the formated data and queries used for efficiency evaluation for example. -
src/
- contains all the files necessary to the implementation of SDEcho.
#Use the following sql to import the data in "data/mini_sample_data" into the database:
copy table_name from file_name with delimiter as ',' NULL '';
#for example:
copy nat2021_01 from 'data/mini_sample_data/nat2021_01' with delimiter as ',' NULL '';
#Configure database information in "src/main.cpp", which are "port", "dbname", "user" and "password".
# compile
cd src
mkdir build
cd build
cmake ..
make
# run
cd src
sh test.sh
#All baselines are compiled and run in the same way as SDEcho.
#Initialize three databases from TPC-H: 'origin' for original data, 'tpch_rq1' for data after insertion of patterns and 'noise' where the inserted patterns are selected from. Notice that databases 'origin' and 'tpch_rq1' share the same scale while the scale of 'noise' should be larger.
#We provide data sets of different scales under "./benchmark/TPC-H".
#Use the following sqls to import data, change the scale size to import TPC-H data at different scales.
copy region from './benchmark/TPC-H/scale_0.001/region.tbl' with delimiter as '|' NULL '';
copy nation from './benchmark/TPC-H/scale_0.001/nation.tbl' with delimiter as '|' NULL '';
copy partsupp from './benchmark/TPC-H/scale_0.001/partsupp.tbl' with delimiter as '|' NULL '';
copy customer from './benchmark/TPC-H/scale_0.001/customer.tbl' with delimiter as '|' NULL '';
copy lineitem from './benchmark/TPC-H/scale_0.001/lineitem.tbl' with delimiter as '|' NULL '';
copy orders from './benchmark/TPC-H/scale_0.001/orders.tbl' with delimiter as '|' NULL '';
copy part from './benchmark/TPC-H/scale_0.001/part.tbl' with delimiter as '|' NULL '';
copy supplier from './benchmark/TPC-H/scale_0.001/supplier.tbl' with delimiter as '|' NULL '';
#Modify the database connection information in the file 'benchmark/RQ1/connect.py'
#generate inserted patterns for a specific sql
python benchmark/RQ1/dbgen.py
More datasets of large scales
can be downloaded from https://drive.google.com/drive/folders/1CDlnt1sLHbh0JJXlH7Y4wyHR2rKCT_QK?usp=sharing
python benchmark/RQ1/explainEngine.py