Link to paper: https://www.sciencedirect.com/science/article/abs/pii/S0957417420304371

In this work we address the Algorithm Selection Problem, i.e., the decision of which algorithm to use from a set of alternatives, given an instance. For the selection, we take into account a given time limit as a parameter. Hence, the learning is based on the anytime behavior of the algorithms. We test this approach over the well known Knapsack Problem.

Dataset

• Knapsack instances
• Features: 21 features for the 15,000 instances. The instance identifier is the index.
• Solver results

For Python, we recommend using pandas to open files:

import pandas as pd
features = pd.read_csv("features.csv")
solver_results = pd.read_csv("solver_results.csv")