World epidemiological data states that primary dysmenorrhea occurs most frequently in women aged 17-24 years. Multinomial Logistic Regression (MLR) is suitable for producing predictive models with a dependent variable (menstrual pain) that has four categories. This study analyzes the predictive model of risk factors for the incidence of menstrual pain (primary dysmenorrhea) in adolescent girls in Pekalongan City. The research design was an analytic survey with a cross-sectional design. A total of 100 samples were collected using multistage random sampling at four school sites. The questionnaires used included NRS, PSS-10, and anthropometric measurements. Of the fourteen variables studied, three variables,family history of menstrual pain, sleep duration, and exercise habits, proved to significantly affect the incidence of menstrual pain in adolescent girls (p-value <0.05). The Multinomial Logistic Regression model produced three logit equations. The independent variables, consisting of age, age of menarche, Body Mass Index (BMI), duration of menstruation, menstrual cycle, family history of menstrual pain, meal skipping, night sleep time, morning wake-up time, sleep duration, exercise habits, cigarette smokeexposure, alcohol consumption, and stress level, could influence menstrual pain by 61.2%.Meanwhile, the other 39.8% was influenced by variables not studied. The accuracy of theclassification table, with a prediction truth rate of 80.9%, explained that respondents with a family history of menstrual pain were 98.319 times more likely to have moderate menstrual pain compared to those without a family history of menstrual pain. Thepredictive modeling has good accuracy.