Project banner

Summary of the results of the Execution Phase 1

In this phase we have defined the deterministic model that will be used for calculating the iota indicator. The main principles have been discussed and the mathematical equations have been defined that will be used iteratively on the nodes of the citation network. Two tests have also been fixed which can be applied in case of the used initial conditions. The algorithm to be implemented has also been established. It was observed that, in order to be more efficient, before implementing the algorithm a topological sorting of the network is required. The first database has also been prepared based on citations from the Web of Sciences.

Summary of the results of the Execution Phase 2

In this phase first we have implemented the deterministic model for calculating the iota indicator on two citation databases: Web of Science (WoS), and High Energy Physics (HEP). First, the databases have been prepared by topological sorting in order to perform an efficient calculation of the indicator values. The obtained values depend strongly on the year of publication which can be detected by a trend which favorizes the old articles. In order to eliminate this trend, analytical studies have been performed on Erdos-Rényi type networks. These studies resulted in the redefinition of the deterministic model. Similar methods have been successfully applied in case of WoS and HEP citation networks which resulted a time-normalized article indicator which is not sensible to the initial conditions.

The main objective in case of the probabilistic model is to estimate the precision of the indicator values, as well. In this model the indicator value is a random variable with unimodal distribution, and the master equations are written by convoluting the probability distribution functions. However, computig convolutions on very large networks are very costly. For our purpose it would be enough to estimate the standard deviation of the article indicator values ony. We have shown that in case of exponential distribution functions  these standard deviations can be efficiently calculated. Currently we are working on the implementation of this efficient algorithm.

Summary of the results of the Execution Phase 2

In this phase we have developed the algorithms needed to evaluate the level of accuracy of the indicator value in case of the probabilistic model. According to the mathematical model developed in the previous phase, the direct calculation of the accuracy level involved significant computational resources making the process unfeasible. By using a recursive algorithm, however, the accuracy level may be approximated with arbitrary precision and convenient computational costs.

Moreover, a detailed study has been carried out regarding the comparison of iota values to impact factors in case of WoS citation network. The time evolution of the iota indicators have been studied, as well. The calculated indicators show good correlation with impact factors after a long time period since the publication date.

We have conducted a study on the cross-disciplinary normalization of the iota indicator. Is was shown that the indicator is normalized for different scientific domains and therefore, the normalization method should not be applied.

Finally, it was shown that the role of the initial values in the iota algorithm, which can be the impact factor of the journal, article ratings provided by readers or its combination, is important in the case of new articles. However, as time goes these initial values are diminuated and the iota value will be governed only by the citations received from other articles.


Working Phase 1: Completed (Technical report delivered in December 2016)

Working Phase 2: Completed (Technical report delivered in November 2017)

Working Phase 2: Completed