Claim Missing Document

Found 1 Documents

Penerapan Algoritma Regresi Linier pada Prediksi Tarif Influencer Media Sosial Muhammad Riza Alifi; Hashri Hayati; Cholid Fauzi
Journal of Information System Research (JOSH) Vol 4 No 1 (2022): October 2022
Publisher : Forum Kerjasama Pendidikan Tinggi (FKPT)

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (764.283 KB) | DOI: 10.47065/josh.v4i1.2361


The influencer industry has emerged as a result of social media disruption, and its members can affect audience interest in the goods and services being advertised. Because advertising performance on social media is more quantifiable than it is with traditional media, using influencer services is thought to be preferable. Influencer rates are often dependent on reach, engagement, and follower count. However, since there is no reference standard used in determining the prices, it could harm one of the parties. In order to reduce the impact of losses for both influencers in giving rates and clients in accepting rate offers, this study intends to propose a solution in the form of a machine learning-based influencer rate prediction model that can be used as a reference. The stages of this study are literature review, data gathering, pre-processing of the data, linear regression model development, and model evaluation. Five different models were produced as a result of this investigation. One of the best models has an MAE of 145401.484375, an MSE of 7.222241e+10, and an RMSE of 268742.250. These findings are affected by the hyperparameter learning rate of 0.001 and the epoch of 1,000. Most of the test data have not been completely represented by the model. The little number of datasets utilized for training, only 161 rows with 4 positively correlated attributes, is one of the reasons why the model is not really optimal. Nevertheless, from the standpoint of using a relatively small dataset, the model developed in this study is quite successful because several of the prediction results are fairly near to the real value, one of which is the prediction value with an error difference of −347.69.