| _version_ | 1866901544600862720 |
|---|---|
| author | Yazan Aswad Dr. Aghiad Kh. Alkatan |
| author_facet | Yazan Aswad Dr. Aghiad Kh. Alkatan |
| contents | Traditional models for predicting future sales of a product or service are based on previous, not updated data, resulting in unsatisfactory and inaccurate forecasting results, meaning that the data used as inputs to the forecasting process is stable and not dynamic during the forecasting process. The research aims to leverage social media data by extracting features from Facebook platform (features are reactions to posts) and using them as input to the automated forecasting system to try to predict corporate revenues. Machine learning algorithms have been trained to predict returns according to pre-stored data and can be updated on demand, which means that the proposed forecasting system will work in a dynamic environment. The following algorithms were used to predict the profitability of new services and the one with the highest accuracy was selected: (Random Forest, DT, Gradient Boosting, K nearest neighbors, NB). The results showed that Random Forest algorithm is the one with the best accuracy, with an accuracy of 67%, and a slight correlation was observed between the interactions on the target post and the profitability of the service within the post. |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_18439332 |
| institution | Zenodo |
| language | |
| publishDate | 2022 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Using Social Media Data To Forecast Telecom Companies Revenues with Machine Learning Yazan Aswad Dr. Aghiad Kh. Alkatan Machine learning social media marketing classification (Random Forest DT K nearest neighbors NB) sentiment analysis Facebook graph API. Traditional models for predicting future sales of a product or service are based on previous, not updated data, resulting in unsatisfactory and inaccurate forecasting results, meaning that the data used as inputs to the forecasting process is stable and not dynamic during the forecasting process. The research aims to leverage social media data by extracting features from Facebook platform (features are reactions to posts) and using them as input to the automated forecasting system to try to predict corporate revenues. Machine learning algorithms have been trained to predict returns according to pre-stored data and can be updated on demand, which means that the proposed forecasting system will work in a dynamic environment. The following algorithms were used to predict the profitability of new services and the one with the highest accuracy was selected: (Random Forest, DT, Gradient Boosting, K nearest neighbors, NB). The results showed that Random Forest algorithm is the one with the best accuracy, with an accuracy of 67%, and a slight correlation was observed between the interactions on the target post and the profitability of the service within the post. |
| title | Using Social Media Data To Forecast Telecom Companies Revenues with Machine Learning |
| topic | Machine learning social media marketing classification (Random Forest DT K nearest neighbors NB) sentiment analysis Facebook graph API. |
| url | https://doi.org/10.5281/zenodo.18439332 |