This project is made for hackathon HACKJAIPUR in which we are predicting the mortality rate of hepatitis patients with the help of given dataset.
Keyword search on the Internet has become one of the main behavioral characteristics of human understanding of Hepatitis B, with a prediction accuracy that can be significantly improved through the use of a comprehensive set of online search indices. The main objective of this study is to establish and apply a prediction model for Hepatitis B incidence, in order to provide a scientific basis for Hepatitis B early detection and timely remediation. Firstly, we utilized principal component analysis to extract salient information. Next, we established a comprehensive prediction model for Hepatitis B trends using the stepwise regression method. Finally, we established a time series model and search data model for comparison, and also predicted the incidence of the next period. The results showed that the model provides stable and timely data, which makes it an ideal tool for prediction, while also providing a frame of reference for the prediction of other infectious diseases.
Early and effective prediction of the incidence trend of Hepatitis B can provide a scientific basis for the prevention and treatment of Hepatitis B, as well as for the rational allocation of health resources, thereby reducing unnecessary waste. However, existing office Hepatitis B testing organizations in almost every country only collect the number of suspected Hepatitis B cases in the hospital population to use it as the survey data on Hepatitis B incidence. This approach requires the establishment of a nationwide monitoring network, where the collection and processing of data have to go through complex processes. As a result, the monitoring data for lags behind the actual development of the disease [4]. In light of this, there are direct prediction models of Hepatitis B incident trends based on time series, combined model, and the gray model. These studies help to improve the accuracy and timeliness of short-term prediction mechanisms.
At present, researchers at home and abroad are actively using a variety of methods to come up with different prediction schemes, but most of them are based on short-term data for correlation and extrapolate the future using historical data. Although such methods also have some practical significance, the resulting schemes tend to be untimely in their predictions, which are especially problematic for populations of infected persons that can change suddenly and dramatically within very short periods of time. In addition, the analysis of a single factor is not enough to be able to fully grasp the characteristics of epidemics and laws of infectious diseases, and the application conditions of the various models differ. The accuracy may differ when using different prediction models for fitting the incidence of the same infectious disease [20]. Currently, a variety of methods are being used in combination to study the pathogenesis of Hepatitis B. For example, systematic reviews, meta-analysis of literature databases, DNA big data analysis, and clinical trials are being used to understand the pathogenesis of Hepatitis B from a multi-angle view [21]. This research attempts to come from a completely new perspective and is a useful complement to traditional research methods.
In recent years, the Internet has been integrated into the field of smart medical care and now plays an important role in the prediction and prevention of many diseases. For example, Google Trends have proved useful in the prediction of certain diseases for a specific term [22]. Scholars have facilitated the faster and more accurate prediction of syphilis by investigating the relationship between online search behavior and the actual number of diseases [23].
With the rise of big data and the massive health industry, a large number of medical information resources are being digitized and platformized. Innumerable patients, medical institutions, and medical personnel are being connected, people who are keen on Internet health consultation and information sharing. And Internet search engines have became the backbone of these platforms as well as served as a bridge between doctors and patients. Therefore, scholars have a strong research interest in health search behavior. Some studies have found that about half of patients still cannot accurately comprehend a diagnosis when leaving the doctor's office, making online patient education sites a major source of information for many patients. Through the Google Search Index, we found that reading materials about Hepatitis B and C, cirrhosis, and liver cancer are mainly focused on the risk factors, symptoms, diagnosis, treatment, and prevention [24]. It can be seen that more and more people are relying on the Internet for health counseling and services. They use the Internet not only to look for health-related information, but also to read more professional medical literature or relevant materials, paying the most attention to measures that can improve life quality. The study found that the more serious the patients' condition, the more frequently the Internet search behavior can lead to a positive understanding of their own health. However, in the previous studies, the impact of Internet use on their own health awareness has not yet been taken seriously. For example, when seeking health advice online, quality, personality, fairness, and credibility of the information enhances online trust and make people willing to take action based on online advice [25]. In fact, patients may selectively manage information from the Internet to form a self-serving and positive perception of their own health, which psychologically improves patient's ability to cope with the disease [26].
In short, predictive methods based on web search data are still in their infancy, and there are still many areas to be explored. For instance, the reasonable number of search terms and the choice of accurate positioning of the number of delay periods have yet to be explored. On this basis, big data forecast modeling of sudden illnesses such as influenza, Hepatitis A, and Hepatitis C is also a key research topic going forward.
*streamlit (https://docs.streamlit.io/en/latest/)
-pip install streamlit
-streamlit run [filename]
| Folder | Description |
|---|---|
| data | Dataset for performing exploratory data analysis (so as to gain data understanding) and for building classification models of the hepatitis patients. |
| models | classification models trained on of the hepatitis dataset. |
| Jupyter Notebook | Contains detailed EDA ,Modelling,Training and Predicting. |
| Streamlit app | Code for streamlit app for predicting hepatitis |